Chemical Engineering and Processing 48 (2009) 892–901

Contents lists available at ScienceDirect

Chemical Engineering and Processing:

Process Intensification
journal homepage: www.elsevier.com/locate/cep

Retrofit of sour water networks in oil refineries: A case study

Daniel Sujo-Nava, Lisa A. Scodari, C. Stewart Slater, Kevin Dahm, Mariano J. Savelski ∗
Rowan University, Department of Chemical Engineering, 201 Mullica Hill Rd, Glassboro, NJ 08028, United States

a r t i c l e i n f o a b s t r a c t

Article history: A case study is presented of the retrofit of a sour water network in a petroleum refinery that generates
Received 7 June 2007 581,000 m3 /yr of sour water. Flowrates and contaminant levels of ammonia and hydrogen sulfide were
Received in revised form 1 December 2008 traced for all major water streams in the sour water network. By grouping some process units, a reuse
Accepted 7 December 2008
scheme was found that only generates 280,000 m3 /yr of sour water, saving 83% of freshwater and 52% in
Available online 13 December 2008
energy and CO2 emissions associated with pumping and the heat requirement to scrub out the sour com-
ponents. Additionally, it was found that the remaining water, after treatment in the sour water stripper,
Keywords:
could be reused elsewhere in the plant, if cyanide ions were eliminated. This case study has provided the
Sour water
Waste minimization
foundation for further plant studies to validate the water profile of the proposed scheme.
Petroleum refinery retrofit © 2009 Elsevier B.V. All rights reserved.
Water allocation
Process integration

1. Introduction this type of water network. The latter scope is the matter of this
paper.
Ammonia and hydrogen sulfide are common pollutants in the The Pollution Prevention (P2) hierarchy can be adapted to solve
oil refining industry. These two pollutants are generated in gas this problem [11]. The use of this strategy provides sufficient ele-
form (sour gas) mainly when oil is cracked to produce fuels with ments for analyzing and proposing a solution to the problem. Since
low sulfur content. The contaminants are scrubbed using water or the generation of sour water cannot be avoided by any means, the
steam that becomes sour water. Besides these two contaminants, P2 strategy should be focused first on minimizing the amount of
sour water may contain other pollutants such as phenol, cyanide, material contaminated in processes and then controlling pollu-
aliphatic, and naphthenic acids, etc. [1]. The amount of produced tants’ release to the environment. The same outcome is obtained
H2 S and NH3 depends directly on the nature of crude (sweet or by using the waste minimization hierarchy [12], which is based on
sour), not the process. All unit operations at a refinery that pro- the conceptual approach for process design [13]. Sour water man-
duce this type of water compose the sour water network (SWN). agement must also address economic feasibility and profitability.
Refining sour crude has increased significantly in recent years due The desired goal is to develop sustainable and green production
to oil scarcity and high prices. However, refining operations are processes by designing systems that optimally satisfy engineering,
constrained economically by the difficulty of meeting environmen- environmental, and economic (E3) objectives. It must be high-
tal regulations. Managing sour water efficiently is imperative for lighted that the interactions between the levels of the hierarchies
mitigating the environmental impact of sweet oil reserves deple- must be dynamic, so that the retrofitted water system considers
tion. the interactions between prevention, reuse, treatment and disposal
Sour water is treated to reduce its NH3 and H2 S concentra- (Fig. 1).
tions typically using steam desorbers or similar unit operations. The The three aforementioned objectives can be achieved by inten-
most common unit found in refineries for this purpose is called the sifying the use of water in process. Such intensification can be
Sour Water Stripper (SWS). Both substances are then recovered in accomplished by reusing water between units before and/or after
a reusable form [2], i.e. molten sulfur. Sour water stripping devices treating. The possibility of reusing water depends mainly on the
have been widely studied, optimized, and commercialized [3–10]. ability of the considered process units to withstand (un)certain
Much research has focused on improving this end-of-pipe solution pollutant levels, their geographical location within the plant, and
for the SWN, whereas less attention has been on optimizing solely the capacity to regenerate their process effluent, while maintain-
ing the operation quality within acceptable standards. The number
of reuse schemes that could be generated is large and each of
them would have a unique E3 performance profile. In quantita-
∗ Corresponding author. Tel.: +1 856 256 5317; fax: +1 856 256 5313. tive terms, objective functions describing the E3 goals can be set
E-mail address: [email protected] (M.J. Savelski). such as minimization of freshwater consumption, operating costs,

0255-2701/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.cep.2008.12.002
 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901 893

Fig. 1. Use of P2 and Waste Minimization hierarchy for process intensification towards sustainability.

network complexity, regeneration costs, pumping capacity, piping presented a set of heuristic rules to design water networks that
requirements, maximization of return of investment, etc. As a result, obtains sub-optimal solutions and sometimes the minimum fresh-
a model can be set up to design efficient water systems. [14] water consumption. However, this type of methods may provide
Several methodologies for searching for efficient water net- good solutions without intricate methodologies.
works have been developed for more than 25 years. A wide range of Graphical tools were initially developed by Wang and Smith [19]
proven tools for designing and retrofitting water systems has been for single and multiple contaminants. This prime method became
presented and used in several case studies. In general, searching very popular because of its simplicity and ease of application.
for an optimal reuse scheme can be approached by heuristics or However, several features of water systems were not addressed
systematical methods that are based on conceptual design insights by this method, such as forbidden matches, profitability, multiple
[13]. The conceptual methods are supported by graphical and/or contaminants, or fixed flowrate systems, regeneration, uncertainty,
mathematical tools [15–17], which would provide a scheme with etc. [20]. Several improved methods were presented to overcome
better E3 performance. Fig. 2 shows a classification by type of the some of these problems [21]. The graphical methods are not robust
design methodologies available. for grassroots designing or retrofitting multi-contaminant sys-
Very few heuristic methods have been presented in the open tems, which are typical in refineries. The generated scheme by the
literature because there is no guarantee of optimality. Liu et al. [18] graphical techniques does not consider other objective functions

Fig. 2. Design methodologies outlook.

894 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901

or all variables at once (piping options, distances between units, relationships between the costs of regeneration and freshwater
etc.). Cost optimization for example, it is not addressed directly by on zero water discharge solutions. The authors used two different
the graphical methods when generating the water-using network. methods depending on the number of contaminants. For the single-
Instead, economic evaluation tools are used to screen the generated contaminant case, they applied the optimality conditions of waters
options that meet the targets. Although it had been argued that schemes [38]. For the multi-contaminant one, a combinatorial
pinch-based methods could be helpful to find an initial solution search along with necessary conditions of optimality was utilized
prior to using mathematical optimization [22], recent evidence in order to solve the problem using linear models [39]. Bagajewicz
indicates that the use of the pinch method may mislead the math- et al. [40] showed the application of mathematical programming
ematical optimization solution [23]. However, a novel approach and the necessary conditions of optimality [41] for the retrofit of
presented by Alva-Argáez et al. uses the Water Pinch projection multi-contaminant water systems. The method can guarantee the
scheme for the development of conceptual decompositions of selection of the global optimal solution (between apparent optimal
mixed integer linear and non-linear programming formulations solutions) featuring minimum capital and operating costs. Faria and
[23–25] in order to develop a water network layout. Bagajewicz [42] discussed in detail the methodology for retrofit of
The mathematical programming methodologies can minimize water networks using mathematical optimization.
the total water flowrate requirements, the interconnections for Despite the rigorousness of the solutions obtained using the
water reuse, and the size of regeneration units, while enhancing mathematical techniques or the pinch method (in some cases),
economical and environmental performance. Such techniques are reuse schemes can be developed by using a priori rules. Such rules
based on the construction of a superstructure, which is a mathe- are logical design concepts that help simplifying the system under
matical representation of all possible interconnections between the consideration and guarantee the reduction of cost and mass/energy
considered process units using material balances [26]. The inlet and intensity of water usage in sour water networks. These rules are
outlet concentrations and flowrates can be constrained depend- applicable for the design of water utilization networks, but particu-
ing on the design of each process unit. An objective function is larly useful in retrofit projects. They provide a quick overview of the
then set, i.e. minimization of the sum of all flowrates. However, possible solutions and enable the practitioner engineer to discern
the complexity of the problem and its solution depends on the the potential of water reuse in financial and environmental terms
number of variables taken into account. Such variables are the without the use of the complicated mathematical techniques or
contaminants in the streams, the number of process and regen- specialized software. These concepts are explained in the following
eration units, fixed mass-transfer or flowrate, operating costs [27], section.
profitability, and their related uncertainty [28,29]. Mathematical Some case studies have been presented in the open literature
optimization techniques can tackle a higher number of variables for the oil refinery industry. Al-Redhwan et al. [28] proposed a
at once, i.e. multiple-contaminant concentrations, limited only by water minimization model, which included ranges of uncertain sea-
the nature of the optimization tool used. The mathematically opti- sonal flowrates and loads. The model was applied to a 400,000
mized networks need to have the least number of interconnections barrels/day throughput refinery located in Kuwait. The optimized
for a safe operation and control (“keep it simple” concept) to be network included an atmospheric residue desulfurization unit, a
accepted by management. Furthermore, the interactions between vacuum distillation/rerun unit, a tail-gas treatment unit, an atmo-
heat exchanger integration and water reuse schemes must be con- spheric residue desulfurization unit, a kerosene desulfurization
sidered as well in order to generate optimal process networks with unit, a fluid catalytic cracking unit, and a desalting unit. The contam-
efficient energy use. Most methods assume the system to be isother- inants considered were ammonia, chlorine, hydrogen cyanide, and
mal, which could have a negative impact on the equipment such as hydrogen sulfide. A cost function was included to assess economic
crystallization of species in piping or units as well as on energy impact. A 58% reduction of freshwater intake was achieved by incor-
use in unit operations. A novel mathematical approach to inte- porating reuse and regeneration/reuse operations. The authors also
grate this important feature of such schemes has been published found that wastewater containing hydrogen cyanide cannot be
recently. Leewongtanawit and Kim [30] proposed an optimization reused because it would cause severe corrosion in other units.
model that is capable to generate water reuse and heat recovery If the cyanide compounds are removed from the sour water, the
networks with multiple contaminants using a decomposition of a reduction of freshwater consumption on this type of water net-
mathematical programming formulation that is constrained by the works can be reduced even further. Few technologies have been
water superstructure modeled as MINLP [31], a heat transshipment developed for this purpose, such as bipolar trickle towers or by
model [32], and the Heat Exchanger Network hyperstructure [33]. reacting hydrogen cyanide with ammonium polysulfide. The for-
The method is capable of devising a scheme with non-isothermal mer technology is a low cost electrochemical reactor investigated
mixing points or a network with a separate HEN. The latter case by Ogutveren and Koparal [43], who carried out experiments using
provides simplicity to the design as well as its cost-effectiveness by Raschig rings as electrodes. Cyanide is collected at the anode and
considering capital costs as well. destroyed through direct oxidation with a strong alkaline solution
Specifically for the retrofit case, several rigorous methodologies (pH 12). If the cyanide is complexed with any heavy metals, they
have been introduced using mathematical programming. Huang are collected at the cathode before the oxidation reaction. Another
et al. [34] proposed a non-linear programming (NLP) formulation, technology for the destruction of this corrosive compound involves
based on the work of Galán and Grossmann [35], that searches for its reaction with ammonium polysulfide, which can be injected
efficient effluent-treatment systems. Joedicke and co-workers [36] on-line or generated inline by oxidizing an aqueous solution that
presented a mixed integer linear programming (MILP) formulation contains ammonium ions and sulfide ions in an electrochemical
that utilizes easy to obtain information such as process location and cell [44]. A continuous in-line polysulfide injection process could
current water demand, as well as pre-made decisions on whether be implemented on the catalytic cracking unit effluent. No commer-
the reuse of water is allowed or not. These authors claimed that gen- cialization of this technology of the scale described in this project
erating schemes in this way would be more likely to be accepted was found.
by management. The main objective of this MILP is minimization Zbontar and Glavic [45] presented a case study of a petrochemi-
of total costs, which includes operating (freshwater, wastewater cal complex and a refinery. For the latter, three lines of atmospheric
treatment, pumping) and investment costs (piping, holding tanks), distillation, and three water treatment operations were considered.
for a given time horizon by screening reuse schemes generated The contaminants measured were wastewater flowrates, tempera-
using a matrix to find matches. Koppol et al. [37] investigated the tures, pH, and pollutants contents. It was found feasible to reduce
 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901 895

the overall fresh water consumption by 7%. The payback period washing purposes (user/producer). However, some sour conden-
was of 6 months for the petrochemical site and 11 days for the sates are also generated within the users. All sour water streams are
refinery. However, it is not explained in depth how the proposed merged and then treated in a tray-column steam desorber (SWS).
flow diagrams were obtained. Considered were pipeline dimen- Finally, molten sulfur is produced and recovered in a sulfur recovery
sions, treatment units, sources and sinks, and costs. A reduction unit downstream.
of 9% of the overall freshwater withdraw was estimated. Linnhoff Fig. 3 shows the relative location of the units to each other and
March Energy Services of KBC Process Technology Ltd. worked the current stream locations. Sour condensate streams are denoted
with Amoco’s oil refinery in Yorktown, VA, after the refining com- by long/short dashed line and washing water streams with a solid
pany negotiated a special arrangement with the EPA to pilot test line. All inlet and outlet streams were characterized as well as
a co-operative approach to environmental compliance. The objec- the operating conditions of the considered process units using the
tive was to accomplish air and water emissions reduction through described methodology in the following section.
process integration instead of using conventional end-of-pipe solu-
tions. Freshwater consumption and wastewater effluent flow were 2.3. Data gathering or water characterization
diminished by 14% and 24%, respectively [46]. Duyvesteijn [47] pre-
sented an overview of the feasibility of recycling treated wastewater To apply any water minimization methodology, all inlet and out-
in an oil refinery. It was found that such recycling is technically fea- let streams need to be characterized. This step is crucial because
sible but financially expensive to be implemented without subsidy. the possibility of reuse is restricted by water quality and accu-
However, the costs could be cut down if a reuse scheme were found. racy of the balances. First, inlet and outlet flowrates were obtained
Alva-Argáez et al. showed the application of their pinch method within ±10% accuracy for all major water streams for every unit in
decomposition methodology to retrofit problems in refineries [24]. the sour water network [48,52,53]. At this point, infeasible outlet
For a further detailed explanation of each method presented in streams were recognized, leading to merging streams or removal
the open literature, the reader is referred to the comprehensive from model. The criterion of such infeasibility was flowrates below
review presented by Bagajewicz [48]. 10 dm3 /min. For instance, one can notice in Table 1 that Producer
In this paper, a case study of the retrofit of an oil refinery SWN to 9, with a flowrate of 3.8 dm3 /min, was removed from the system
reduce freshwater consumption and sour water generation is pre- to simplify the retrofit problem. In addition, streams with similar
sented. The applied methodology is based on heuristics that were contaminant content and nearby location should also be merged
developed according to the demands of management in order to (equivalent piping distances larger than 600 m). However, the con-
show potential savings of implementing a retrofit project without taminants should be selected beforehand.
the use of rigorous methods that may be cumbersome to apply The main pollutants that restrict the reuse of water were
especially at early stages of design. Therefore, the proposed changes selected. It is recommended to select few important contami-
to the network would be accepted for further investigation and nants depending on the type of tool that is going to be applied,
implementation. because with each added contaminant, the complexity of the prob-
lem increases by a geometrical factor. On the other hand, excluding
2. Methodology key contaminants could cause severe operational problems with
process units [54]. For the case presented in this paper, only three
The present study was carried out at a refinery in the North- pollutants were taken into account. Since most of the operations
east of the United States, using the unique “engineering clinics” in the SWN are sour gas scrubbing units, NH3 and H2 S are the
model that Rowan University’s College of Engineering has estab- obvious constraint species to be traced. In addition, the presence
lished. This approach facilitates the partnership between academia of cyanide ions prevented the reuse of water of streams contain-
and industry in solving real-world challenges [49–51]. The refin- ing this pollutant to avoid corrosion problems. The hypothesized
ery’s engineering teams provided the necessary information and corrosion mechanism starts when hydrogen sulfide reacts with the
data to characterize the process and water network. iron on the pipe. A protective layer of FeS is formed but it is dis-
solved by cyanide, exposing the pipe for more corrosion [55–59].
Al-Redhwan et al. [28] proposed that cyanide concentration should
2.1. Problem statement
be zero for reuse. For this reason, water coming out from User 2
cannot be reused. Although aliphatic and naphthenic acids and phe-
For this case study, it is desired to determine the intercon-
nol [1] should also be considered because of fouling in some units
nections needed to reduce sour water generation in an existing
[55,60], these were excluded from this analysis.
SWN in order to assess the benefits of using an advanced math-
Quantifying the limiting pollutant concentration profile is not
ematical optimization tool. The system consists of two users, two
trivial and it should be performed carefully. Establishing a higher
users/producers, and two producers of sour water. Flowrates, con-
taminant loads, freshwater quality, and maximum inlet and outlet
concentrations for all units are given. The effects on the SWS perfor- Table 1
mance due to the resulting changes on the inlet stream composition Water balances for units considered.
were estimated by basic energy balances. Unit Flowrate [dm3 /min]

In Out
2.2. Description of sour water network
User/Producer 1 114 114
User 2 360 360
This refinery is capable of refining 31,000 m3 /day of sour User/Producer 3 151 246
crude (Arab Light and Heavy, Oriente, Kirkuk) and generates User/Producer 4a 19 19
approximately 1105 dm3 /min of sour water. The major sour water User/Producer 4b 45 45
Producer 5 0 19
contributors are the fluid catalytic cracker, the hydrodesulfurization Producer 6a 0 151
units, and the coker units. Ten units, five of which are fed by steam Producer 6b 0 38
(producers), comprise the sour water network and their conden- Producer 7 0 76
sates are sent to the SWS. The remaining units are fed by water from Producer 8 0 38
User/Producer 9 4 4
the cooling tower make-up outside of the sour water network for
896 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901

Fig. 3. Operations currently feeding the sour water network.

limit may lead to equipment fouling [54], which would outweigh its were further restricted for each particular unit based upon the
the fact that more water could be reused. On the contrary, if a ability of the unit to withstand a certain contaminant level, which
lower limit were set, the obtained network scheme would not reuse was evaluated by the clinic and the refinery teams. In addition, the
water at its full extent and therefore water savings would be rela- variation of the contaminant levels was considered, specifically the
tively small. At this stage of the project, plant personnel expertise high peaks historically observed.
was required to assess the impact of using different water qualities Since the units that compose the SWN are mass-transfer based,
[52]. Analyzing the contaminant concentrations of the system gave the pollutant concentrations were utilized to calculate the contam-
insight into the real limit for the maximum outlet concentration inant mass load of every unit. Eq. (1) was formulated to calculate
values. Table 2 shows the supplied concentration data for each of the actual mass load of all users/producers.
the units being considered. The maximum outlet concentration lim-
Lj,s = Fj,out Cj,s,out − Fj,in Cj,s,in (1)

If the mass load Lj,s is maintain fixed, the restriction of the inlet
Table 2
concentration would depend on the quality and quantity of water
Actual Inlet and Outlet flowrates and concentrations after merging.
that leaves the unit [34]. The solubility of the sour mixture is usu-
Unit No. Flowrate (kg/s) Actual concentration [ppm] ally recommended as a starting point [48]. NH3 and H2 S are highly
In Out Ammonia Hydrogen sulfide soluble in water, but their solubility is a function of pH, tempera-
In Out In Out
ture, and vapor pressure [61]. The solubility was assumed constant
for this analysis [62], but set at low values (equal or less than
User 1 1.89 1.89 40 142 10 105
50,000 ppm). The maximum inlet concentration is defined by Eq.
User 2 5.98 5.98 40 3500 10 4890
User 3 4.40 4.40 40 47176 10 54597 (2). Table 3 depicts the values of the considered maximum outlet
User 4 1.07 1.07 40 350 10 300 concentrations and calculated maximum inlet concentrations for
Producer 5 Steam 0.32 N/A 100 N/A 125 each unit.
Producer 6 Steam 3.15 N/A 1247 N/A 1528
Producer 7 Steam 1.26 N/A 235 N/A 300 Fj,out Cj,s,out,max − Lj,s
Cj,s,in,max = (2)
Producer 8 Steam 0.63 N/A 4500 N/A 5112 Fj,in
 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901 897

Table 3 • Solubility of traced species was assumed constant but set at a low
Limiting water quality.
value of 50,000 ppm. Although these conservative values would
Unit No. Maximum outlet concentration limit the amount of water reused, they provide a good starting
Ammonia Hydrogen sulfide point for evaluating the water reduction potential. In addition,
ppm ppm the resulting scheme using such values would ease their approval
User 1 142 105
by management.
User 2 10000 10000
• If the ability of a process unit to withstand a pollutant level that
User 3 50000 50000 is below than any other available outlet stream, such process unit
User 4 1000 1000 should be discarded as a possible user (User/Producer 1). Con-
Producer 5 100 125
trarily, small flowrates with high concentrations of ammonia and
Producer 6 1247 1528
Producer 7 235 300 hydrogen sulfide that no other unit could operate at that level,
Producer 8 4500 5112 those streams should be avoided (Producer 9).
• Process units with similar contaminant concentrations and water
use (user or producer) in the same region should be combined
Pollutant concentrations or flowrate variations of process water into one single stream (User/Producer 4a and 4b combined to
occur due to, for example, season or feedstock changes and have a Use/Producer 4). This action decreases the pumping and piping
major impact on the integrated network reliability. Although this requirements as well as their associated cost.
issue can be addressed using stochastic optimization techniques • If process units are isolated at the furthest region of the plant,
[28,29,63,64], such variations were discussed with the refinery’s these units should be combined in order to reduce pumping costs
personnel and assumed ‘typical’ conditions for setting up limiting (Producers 6a and 6b streams combined to Producer 6).
outlet concentrations. For instance, if the contaminant load of a pro- • Process units in the same relative location should be consid-
cess unit had an uncertain value within a range, the first quarter or ered first for reuse before supplying water from units further
the median were used to set the limit of the water quality entering away to reduce pumping and piping costs. This would ensure
the system. minimal costs, minimal greenhouse gas emissions, and maximal
Some process units may have the potential of reducing their network simplicity for the performance of the water network
water requirements. For this reason, the mass-transfer coefficient only. However, a tradeoff exists if the energy requirements of the
and flowrate conditions were recalculated using the proposed sol- downstream regeneration units are based on volume.
ubility of ammonia and hydrogen sulfide.
As already noted, streams coming from a nearby location were 2.5. Evaluation of economic and environmental performance
merged with those having similar water quality in order to reduce
the complexity of the design problem. In this case, effluents from The operation costs of the proposed network were analyzed
producers 4a and 4b were combined because they have the same and compared to the actual arrangement in the plant. The energy
concentration of both contaminants. Effluents from producers 6a for distributing the water through the network is one of the two
and 6b were joined because they are both located at the far extreme major contributors to the operation cost. For calculating the pump-
area of the plant. These changes simplified the complexity of the ing costs, the assumptions of an economical flow linear velocity
sour water network and reduce pumping costs. The flowrates of of 2 m/s and an arbitrarily fixed energy price of US$9.5/GJ (coal)
the merged streams j were calculated by adding the flowrates of were considered. In addition, greenhouse gas emissions associated
the considered streams, as depicted in the first part of Eq. (3). Simi- to the generation of the pumping energy were estimated using the
larly, the resulting concentrations of each considered contaminant conversion factor shown in Eq. (4) [65].
s streams j were calculated using the second part of Eq. (3). 
⎧  ⎫ 24.62 kgC Oxidized Coal 44 gCO2 1.0 ton
⎪
⎪ Fj = Fi ⎪
⎪ GJ
0.99
Coal Input 12 gC 1000 kg
⎨ ⎬
i
 ∀j ∈ M, ∀s (3) tonCO2
⎪
⎪ c F ⎪
⎪ = 8.94 × 10−2 (4)
⎩ Cj,s = i i,s i ⎭ GJ
Fj

The merged streams were matched to the conditions established The second major contributor to the operation cost is the energy
in Eq. (2). New flowrates and outlet concentrations were calculated. used in the SWS to desorb the sour compounds by heating up the
The combination and allocation of streams were performed using water coming out from the water network. If the throughput of
the heuristics presented in Section 2.4. the unit is reduced, so is the required heat to reach the opera-
tion temperatures. The energy usage of the reboiler of the SWS was
2.4. Heuristics estimated using basic energy balances. The temperature conditions
found for the sour water stripper configuration at the refinery were
Water allocation rules have to be applied, following the eval- used for estimating the heat consumption of the proposed scheme
uation of the resulting reuse schemes in terms of operational in this work. Fig. 4 shows a block diagram depicting the streams
costs, process reliability, and environmental performance (sav- considered for estimating energy use at the reboiler. Eq. (4) and the
ings in energy and greenhouse gas emissions). Merging streams as price of coal were used to calculate the operation cost and GHG
described before provides a system simple enough to find a solu- emissions.
tion by inspection. A summary of the considerations used when
developing the reuse scheme is presented next. 3. Results and discussion

• Any process unit with known concentrations of cyanide in the The necessary interconnections to reduce the amount of sour
effluent is not a candidate for reuse due to corrosion concerns water generated in the considered retrofit case were found by using
(User 3). This action minimizes the impact on maintenance of the guidelines discussed in the previous section. The new water
equipment and pipelines by avoiding the expense of replacing reuse scheme shown in Fig. 5 reduces the flowrate of cooling tower
any piece of equipment. make-up water fed to the sour water network by 84%, consuming
898 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901

only 114 dm3 /min in contrast to the original 689 dm3 /min. The new
process scheme reduces the amount of effluent from the SWS being
sent to wastewater treatment by 56% (1105 dm3 /min reduced to
481 dm3 /min). Table 4 depicts the new inlet/outlet flowrates and
concentrations.
Water reuse focused on the two units with the highest volumet-
ric requirements (User 2 and 3) in order to achieve a large reduction
in both water consumption and sour water generation. The remain-
ing units were grouped by their location in order to avoid abrupt
changes or expenses that would jeopardize the development of this
project.
More water could be reused by dividing a stream with highly
concentrated sour water into smaller streams, and then matching
these divisions to users within the specified maximum inlet con-
centrations. However, a tradeoff exists when the flowrates of such
type of division are too small to be economically feasible. This was
the case for Producer 8, whose outlet stream was not reused because
its high pollutant concentration.
Fig. 4. Material and energy balance envelop for estimating reboiler workload to the By revisiting the material balances of every unit, it is possible
sour water stripper at different operating hydraulic flowrates.
to find that some mass-transfer rates are not used fully. In this
case, the flowrate requirements were downgraded for User 2 due to
its high contaminant level tolerance and the flow requirements for

Fig. 5. Proposed reuse scheme.

 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901 899

Table 4
Obtained conditions for retrofitted system.

Unit No. Flowrate (kg/s) Concentration [ppm] Load NH3 (kg/s) Load H2 S (kg/s) Flow out NH3 (kg/s) Flow out H2 S (kg/s)

In Out Ammonia Hydrogen sulfide

In Out In Out

User 1 1.89 1.89 40 142 10 105 1.92E−04 1.79E−04 2.67E−04 1.98E−04

User 2 3.46 3.51 1143 7018 1400 9690 2.06E−02 2.91E−02 2.46E−02 3.40E−02
User 3 3.15 4.82 282 30196 284 23601 1.44E−01 1.12E−01 1.46E−01 1.14E−01
User 4 1.26 1.26 235 498 300 546 3.31E−04 3.10E−04 6.27E−04 6.88E−04
Producer 5 Steam 0.32 0 100 0 125 1.89E−05 3.62E−05 3.15E−05 3.93E−05
Producer 6 Steam 3.15 0 1247 0 1528 3.92E−03 4.81E−03 3.50E−04 3.46E−04
Producer 7 Steam 1.26 0 235 0 300 2.96E−04 3.77E−04 2.96E−04 3.77E−04
Producer 8 Steam 0.63 0 4500 0 5112 2.83E−03 3.22E−03 2.83E−03 3.22E−03

Table 5 in the United States. Although the savings are low in comparison to
Energy and pollution prevention summary for SWN.
the whole plant, it would contribute to the overall green engineer-
Scheme Current New proposal Savings ing effort, as well as to support the environmental stewardship of
3
Sour water generation (m /yr) 581173 278753 52%
the company.
Freshwater consumption (m3 /yr) 362379 59958 83% Moreover, environmental and economical goals can be achieved
Energy used (GJ/yr) 2043 932 54% by saving material and energy as shown in Tables 5 and 6. The
CO2 emissions (tons/yr) 183 83 54% reduction of flowrate reduces material and consumption, which
Operation cost (US$) 19350 8825 54%
represents savings of 686,407 USD in annual operational costs.
The proposed changes require repiping about 4500 m. The net-
adequate contaminant removal. The outlet flowrate in User 2 was work changes would have a cost of 248,000 USD. However, this
reduced by 41% and the corresponding contaminant levels were investment has a payback time of 131 days.
doubled for both NH3 and H2 S. Despite the conservative and simplistic approach taken in this
The proposed network reaches almost zero freshwater con- work, it may be possible to find a better or an optimal water
sumption. User 1 requires concentrations at the inlet stream to be network. In order to achieve this, detailed studies should assess
lower than 40 ppm and 10 ppm of H2 S and NH3 , respectively. Such accurately the impact of using more highly concentrated water on
concentrations cannot be achieved by reusing water from any other the units as well as the impact of cyanide ions or organic acids.
unit within the SWN. A solution for avoiding the use of 114 dm3 /min Further, the impact of temperature of the streams should also be
of cooling tower water in User 1 is recycling the bottom effluent of taken into account as well as their impact on the integrated heat
the SWS. However, the source of cyanide must be controlled by a exchanger network.
treatment unit at the exit of the generating units. This would allow Since the reduction of the inlet stream flowrate to the sour water
reusing the regenerated sour water elsewhere in the plant as well. stripper was substantial, some adjustments are needed to address
Further analysis on the removal efficiency of the SWS is required to appropriately the mass-transfer active area of the column and the
quantify the possibility of such reuse. hydraulics of the column, i.e. the rate of bottoms recycle to the feed,
The environmental benefits of the reuse scheme are shown active area of trays, feed pressure and temperature, etc. The column
in Table 5. These metrics for the environmental footprint of the will have a better performance due to the increased concentration
above discussed schemes have and error of ±10% to those shown of the inlet stream, and the reduced load to the column will posi-
in SimaPro 7.0 (Life cycle assessment software). Energy consump- tively affect the efficiency of the column to remove H2 S and NH3 as
tion for pumping is the determining factor for achieving economical well as reducing the heat requirements for stripping. These adjust-
savings, as well as reducing CO2 emissions related to the produc- ments were not part of the scope of this paper, but they should be
tion of this energy within the water network. Since the units cannot analyzed in further detail.
be relocated in this retrofit case, water is reused first in the clos- As mentioned in the introduction, the scarcer sweet resources
est unit to avoid increasing GHG emissions and operational cost of oil will require refineries to upgrade their systems for handling
of the network. The only exception was made for User 1, whose sourer material. Such upgrades may require handling a larger vol-
inlet and outlet concentrations restricted reuse of water in this unit. ume of sour water, along with increased treatment capacity. This
The pumping cost of the avoided flowrate in the proposed network case study provides the basis of analysis for reducing the water
arrangement is 10,525 USD. throughput in unit operations that produce or use sour water before
The proposed network reduces the required energy to heat up expanding the sour water treatment capacity. For instance, if the
the sour water at the SWS by 52%. The saved heat avoids the refinery required increasing its sour water treatment capacity by 1.5
generation of approximately 12,000 metric tons of CO2 and the times, the “easiest option” would be installing an extra sour water
expenditure of 675,882 USD in costs. Table 6 summarizes these sav- stripper. The costs of such installation are listed in Table 7. Assum-
ings. It should be noted that the impacts on the currently integrated ing that the capitalization for carrying out the proposed scheme of
heat exchanger network was not assessed. Energy reductions would
have an economical impact if a tax on greenhouse gas were imposed Table 7
Costs of a hypothetical installation of new sour water stripper (in thousands of US
dollars per year).
Table 6
Energy and pollution prevention summary for SWS. Cost USD/yr

Energy input GHG generation Operation cost Installation (linear depreciation) 1000
required (GJ/yr) (tons/yr) (dollars/yr) Utilities 4900
Operating labor 22
Current scheme 1.37E + 05 1.22E + 07 $ 1,298,871
Environmental hidden costs 1303
New scheme 6.56E + 04 5.84E + 06 $ 622,989
Difference 7.11E + 04 6.33E + 06 $ 675,882 Total 7225
900 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901

the sour water network is 2 million USD and avoids the construction [2] Anonymous, WWT Two-Stage Sour Water Stripping, Chevron Products Com-
of the new sour water network, the payback period for this project pany, San Francisco, CA, 1998.
[3] G.A. Melin, J.L. Niedzwiecki, A.M. Goldstein, Optimum design of sour water
would be less than 6 months. strippers, Chem. Eng. Prog. 71 (6) (1975) 78–82.
A similar analysis can be applied to unit operations that involve [4] S.Y. Lee, J.M. Lee, D. Lee, I. Lee, Improvement in steam stripping of sour water
organic pollutants in their streams and that are treated in a common through an industrial-scale simulation, Korean J. Chem. Eng. 21 (3) (2004)
549–555.
regeneration unit, such as a “benzene recovery unit”. [5] R. Thiele, O. Brettschneider, J.U. Repke, H. Thielert, G. Wozny, Experimental
Process or plant engineers can assess the potential for reuse in a investigations of foaming in a packed tower for sour water stripping, Ind. Eng.
simple way by following the short method presented in this paper Chem. Res. 42 (7) (2003) 1426–1432.
[6] G.C. Hoogendoorn, R.D. Abellon, P.J.M. Essens, J.A. Wessenlingh, Desorption of
and determine if a detailed assessment is appropriate. volatile electrolytes in a tray column (sour water stripping), Chem. Eng. Res.
Des. 66A (1988) 483–502.
[7] D. Lee, J.M. Lee, S.Y. Lee, I.B. Lee, Dynamic simulation of the sour water stripping
4. Conclusions process and modified structure for effective pressure control, Chem. Eng. Res.
Des. 80A (2) (2002) 167–177.
Using heuristic rules to simplify the modeling of a sour water [8] G.J. Walker, Design sour water strippers quickly, Hydrocarbon Process. 48 (6)
(1969) 121–124.
network, a reuse scheme was proposed that reduces significantly
[9] M.A. Isla, H.A. Irazoqui, J. Cerda, Improving sour water strippers, Hydrocarbon
the amount of process water contaminated by hydrogen sulfide and Process. 68 (8) (1989) 65–66.
ammonia. After applying the rules, the network was simple enough [10] D.G. Rodriguez, Sour Water Stripper: Its Design and Application, AICHE Sym-
to find a reuse scheme, which shows great potential for implemen- posium Series 70 (1973) 667–670.
[11] D.T. Allen, D.R. Shonnard, Green Engineering, Prentice Hall PTR, Upper Saddle
tation and justifies the use of a rigorous optimization tool for this River, NJ, 2002.
system. [12] J.M. Douglas, Process synthesis for waste minimization, Ind. Eng. Chem. Res. 31
The proposed scheme reduces energy use and its associated pol- (1992) 238–243.
[13] J.M. Douglas, Conceptual Design of Chemical Processes, McGraw-Hill, New York,
lution and operational costs by 52%. There is a direct relationship 1988.
between these major goals when designing water networks. [14] A.W. Westerberg, A retrospective on design and process synthesis, Comput.
Special consideration needs to be taken when handling streams Chem. Eng. 28 (2004) 447–458.
[15] Y.A. Liu, B. Lucas, J. Mann, Up-to-date tools for water-system optimization,
containing significant levels of cyanide with ammonia and hydro- Chem. Eng. -New York 111 (1) (2004) 30–41.
gen sulfide. Cyanides reduce reuse opportunities due to corrosion [16] R.F. Dunn, M.M. El-Halwagi, Process integration technology review: background
problems that occur. In this study, it is suggested that cyanides and applications in the chemical process industry, J. Chem. Technol. Biotechnol.
78 (2003) 1011–1021.
should be treated using polysulfide after the FCC to allow for the [17] M.M. El-Halwagi, Process Integration, Academic Press, San Diego, CA, 2006.
reuse of the sour water stripper bottoms within the network or [18] Z. Liu, Z.J. Zhang, L. Hu, Z. Wu, Wastewater minimisation using a heuristic
elsewhere in the plant. procedure, Int. J. Chem. React. Eng. (2004), Article A25.
[19] Y.P. Wang, R. Smith, Wastewater minimisation, Chem. Eng. Sci. 49 (7) (1994)
Further feasibility studies must be carried out on handling more
981–1006.
highly concentrated sour water in the equipment before the com- [20] R. Smith, Process integration extends its reach, Chemical Processing (2004)
plete implementation of the scheme. In addition, the impact on www.chemicalprocessing.com/articles/2004/359.html (accessed: 1/2/2008).
the integrated heat exchanger network must be assessed given the [21] Y.L. Tan, Z. Manan, A Retrofit of water network with optimization of existing
regeneration units, Ind. Eng. Chem. Res. 45 (2006) 7592–7602.
changes of the stream temperatures due to the proposed scheme. [22] G. Parthasaray, G. Krishnagopalan, Systematic reallocation of aqueous resources
using mass integration in a typical pulp mill, Adv. Environ. Res. 5 (2001) 61–
79.
Acknowledgements [23] A. Alva-Argáez, A.C. Kokossis, R. Smith, A conceptual decomposition of MINLP
models for the design of water-using systems, Int. J. Environ. Pollut. 29 (1/2/3)
The authors greatly appreciate the United States Environmental (2007) 177–205.
[24] A. Alva-Argáez, A.C. Kokossis, R. Smith, The design of water-using systems in
Protection Agency Grant #NP-97275905. They acknowledge Timo- petroleum refining using a water-pinch decomposition, Chem. Eng. J. 128 (1)
thy Moroz and Anthony Furiato from Rowan University, and Debora (2007) 33–46.
Faria from University of Oklahoma for their input and comments on [25] M. Gunaratnam, A. Alva-Argáez, A. Kokossis, J.K. Kim, R. Smith, Automated
design of total water systems, Ind. Eng. Chem. Res. 44 (2005) 588–599.
this work. [26] N. Takama, T. Kuriyama, L. Shiroko, T. Umeda, Optimal water allocation in a
petroleum refinery, Comput. Chem. Eng. 4 (1980) 251–258.
[27] R. Karuppiah, I.E. Grossmann, Global optimization for the synthesis of inte-
Appendix A. Nomenclature grated water systems in chemical processes, Comput. Chem. Eng. 30 (2006)
650–673.
[28] S.A. Al-Redhwan, B.D. Crittenden, H.M.S. Lababidi, Wastewater minimiza-
tion under uncertain operational conditions, Comput. Chem. Eng. 29 (2005)
Ci,s concentration of contaminant s of stream I prior to merg- 1009–10021.
ing (kg/s) [29] R. Karuppiah, I.E. Grossmann, Global Optimization of Multiscenario Mixed Inte-
ger Nonlinear Programming Models Arising in the Synthesis of Integrated Water
Cj,s,in concentration of contaminant s entering process j (mg/kg) Networks Under Uncertainty, Elsevier B.V., Comput. Chem. Eng. 32 (1–2) (2008)
Cj,s,in,max maximum allowable inlet concentration of contaminant 145–160.
s for process j (mg/kg) [30] B. Leewongtanawit, J.-K. Kim, Synthesis and optimisation of heat-integrated
multiple-contaminant water systems, Chem. Eng. Process 47 (4) (2008)
Cj,s,out concentration of contaminant s leaving process j (mg/kg)
670–694.
Cj,s,out,max maximum allowable outlet concentration of contami- [31] A. Alva-Argáez, Ph.D. Thesis, UMIST, Manchester, U.K., 1999.
nant s for process j (mg/kg) [32] S.A. Papoulias, I.E. Grossmann, A structural optimization approach in pro-
cess synthesis: heat recovery networks, Comput. Chem. Eng. 7 (1983) 707–
Fi flowrate of stream i prior to merging (kg/s)
721.
Fj flowrate of merged stream j (kg/s) [33] C.A. Floudas, A.R. Ciric, Strategies for overcoming uncertainties in heat
Fj,in total flowrate into process j (kg/s) exchanger network synthesis, Comput. Chem. Eng. 13 (1989) 1133–1152.
Fj,out total flowrate out of process j (kg/s) [34] C. Huang, C. Chang, H. Ling, C. Chang, A mathematical programming model
for water usage and treatment network design, Ind. Eng. Chem. Res. 38 (1999)
Lj,s mass load of contaminant s in process j (kg/s) 2666–2679.
M total number of merged streams [35] B. Galán, I.E. Grossmann, Optimal design of distributed wastewater treatment
networks, Ind. Eng. Chem. Res. 37 (1998) 4036–4048.
[36] B. Jödicke, U. Fischer, K. Hungerbühler, Wastewater reuse: a new approach to
References screen for designs with minimal total costs, Comput. Chem. Eng. 25 (2001)
203–215.
[1] R.P. Cahn, N.N. Li, R.M. Minday, Removal of ammonium sulfide from wastewater [37] A.P.R. Koppol, M.J. Bagajewicz, B.J. Dericks, M.J. Savelski, On zero water discharge
by liquid membrane process, Environ. Sci. Technol. 12 (9) (1978) 1051–1056. solution in the process industry, Adv. Environ. Res. 8 (2004) 151–171.
 D. Sujo-Nava et al. / Chemical Engineering and Processing 48 (2009) 892–901 901

[38] M. Savelski, M. Bagajewicz, On the use of linear models for the design of water the Conference of the American Society of Engineering Educators, 2004, Session
utilization systems in refineries and process plants, Chem. Eng. Res. Des. 79 1413.
(2001) 600–610. [52] S. Bédard, M. Sorin, C. Leroy, Application of process integration in water re-use
[39] M. Bagajewicz, M. Rivas, M. Savelski, A new approach to the design of water projects, Pulp Pap. -Canada 102 (3) (2001) 53–56.
utilization systems with multiple contaminants in process plant, Annual AIChE [53] D.C.Y. Foo, Z.A. Manan, M.M. El-Halwagi, Correct identification of limiting water
Meeting (1999). data for water network synthesis, Clean Technol. Environ. Policy 8 (2006)
[40] M.J. Bagajewicz, M. Rivas, M.J. Savelski, A robust method to obtain optimal and 96–104.
sub-optimal design and retrofit solution of water utilization systems with mul- [54] P. Zheng, X. Feng, F. Qian, D. Cao, Water system integration of a chemical plant,
tiple contaminants in process plants, Comput. Chem. Eng. 24 (2000) 1461–1466. Energy Convers. Manage. 47 (2006) 2470–2478.
[41] M.J. Savelski, M.J. Bagajewicz, On the necessary conditions of optimality of [55] W.F. Maguire, Reuse sour water stripper bottoms, Hydrocarbon Process. 54 (9)
water utilization systems in process plants with multiple contaminants, Chem. (1975) 151–152.
Eng. Sci. 58 (2003) 5349–5362. [56] S.H. Frisbie, D.K. Nelsen, S.S. Croce, Cyanide generation, corrosion, treatment,
[42] D.C. Faria, M.J. Bagajewicz, Retrofit of water networks in process plants, XXII and discharge at a petroleum refinery, Corrosion 98, 1998, Paper 584, London.
Interamerican Congress of Chemical Engineering, 2006, Paper 98-d. [57] R.D. Kane, M.S. Cayard, S. Srinivasan, R.J. Horvath, Assessment of Corrosivity
[43] B.U. Ogutveren, E. Toru, S. Koparal, Removal of cyanide by anodic oxidation for in Refinery Sour Water Systems, Eurocorr/2000, Institute of Materials, London,
wastewater treatment, Water Res. 33 (1999) 1851–1856. September 2000.
[44] V. Bruecken, G. Ungar, H. Stonner, M. Stoldt, E. Schlauer, J. Anastasijevic, E. Hill- [58] D.C. Baker, C.C. Chou, Cyanide Occurrence and Treatment in the Petrochem-
richs, Process for decreasing the corrosiveness of a sour water, United States ical Industry, Conference on Cyanide and the Environment, December 1984,
Patent 5,431,877 (1995). Colorado State University.
[45] L. Zbontar, L. Glavic, Total site: wastewater minimization. Wastewater reuse and [59] D. Urban, S. Frisbie, S. Croce, Compliance strategy for cyanides in petroleum
regeneration reuse, Resour. Conserv. Recycl. 30 (2000) 261–275. refinery wastewater: part 1—source characterization and treatment, Environ.
[46] R.A. Tainsh, A.R. Rudman, Practical techniques and methods to develop an effi- Prog. 16 (3) (1997) 171–178.
cient water management strategy, in IQPC Water Recycling and Effluent Re-Use, [60] T. Armstrong, Optimize sour water treatment, Hydrocarbon Process. 82 (6)
26th/27th April, Linhoff March International (1999). (2003) 77–79.
[47] C.P.T.M. Duyvesteijn, Water re-use in an oil refinery, Desalination 119 (1998) [61] M.A. Islam, M.A. Kalam, M.R. Khan, Reactive gas solubility in water: an empirical
357–358. relation, Ind. Eng. Chem. Res. 39 (2000) 2627–2630.
[48] M.J. Bagajewicz, A review of recent design procedures for water networks in [62] J.J. Carroll, A discussion of the effect of pH on the solubility of hydro-
refineries and process plants, Comput. Chem. Eng. 24 (2000) 2093–2113. gen sulfide, AQUAlibrium, 1998, www.telusplanet.net/public/jcarroll/ion.htm,
[49] C.S. Slater, M.J. Savelski, R.P. Hesketh, Green engineering design through accessed 21/04/2007.
project-based industrial partnerships, in: Proceedings of the Conference of the [63] R.R. Tan, D.E. Cruz, Synthesis of robust water reuse networks for single com-
American Society of Engineering Educators, Chicago, IL, 2006, Paper 279. ponent retrofit problems using symmetric fuzzy linear programming, Comput.
[50] C.S. Slater, M.J. Savelski, R.P. Hesketh, Academic–industrial partnerships to Chem. Eng. 28 (2004) 2547–2551.
advance pollution prevention, in: Proceedings of the 2006 Conference of the [64] C. Riverol, M.V. Pilipovik, C. Carosi, Assessing the water requirements in refiner-
American Society of Engineering Educators, Chicago, IL, 2006, Paper 279. ies using possibilistic programming, Chem. Eng. Process. 45 (2006) 533–537.
[51] S. Farrell, R.P. Hesketh, C.S. Slater, M.J. Savelski, Industry and academia: a syn- [65] U.S. Climate Technology Cooperation Gateway, Greenhouse gas equivalencies
ergistic interaction that enhances undergraduate education, in: Proceedings of calculator, www.usctgateway.net/tool accessed on: 22/04/2007.