You are on page 1of 15

Chemical Engineering Science 68 (2012) 313–327

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

Targeting and design of energy allocation networks with carbon capture


and storage
Akshay U. Shenoy a, Uday V. Shenoy b,n
a
Narsee Monjee Institute of Management Studies, NMIMS, V.L. Mehta Road, Vile Parle (W), Mumbai 400056, India
b
Synew Technologies, A 502, Galleria, Hiranandani Gardens, Powai, Mumbai 400076, India

a r t i c l e i n f o abstract

Article history: The mathematical formulation for targeting during energy allocation with carbon capture and storage
Received 22 July 2011 (CCS) is formally developed. For operating-cost optimization with zero excess, it is shown that CCS
Accepted 20 September 2011 sources may be regarded as resources with their cost taken as the increment over the non-CCS option.
Available online 29 September 2011
CCS sources along with clean-carbon resources may then be targeted by profile matching with the
Keywords: limiting composite to establish optimal primary cases. The limiting composite curve is itself sacrosanct
Design and obtained by a single computation of the composite table algorithm (CTA) including only non-CCS
Energy sources. Carbon emission networks (CENs) are designed by the nearest neighbors algorithm (NNA).
Optimization A cost criterion is established to determine cost-factor ranges for optimality of the primary cases, and
Systems engineering
results validated by solving linear programming (LP) and mixed integer linear programming (MILP)
Environmental emissions
formulations. The methodology essentially comprises four distinct stages – targeting, network design,
Carbon footprint
cost analysis, and optimization – with the first two stages not requiring any cost data.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction than air is used for combustion), and integrated gasification


combined cycle (IGCC where self-generated hydrogen is utilized
Optimal planning of energy systems subject to carbon foot- for combustion).
print constraints aimed at reducing climate change effects has The objective of this work is to develop an algorithmic
been the focus of recent research (Tan and Foo, 2007; Foo et al., procedure with a graphical representation for the carbon-con-
2008; Lee et al., 2009; Pekala et al., 2010; Shenoy, 2010). This is in strained energy planning problem with CCS that will identify an
direct response to policies on sustainable development, including optimum strategy for integrating sources and demands to mini-
the Kyoto Protocol, that attempt to balance the expected increase mize operating cost. CCS may be simply viewed as a means to
in energy demand and the desirable decrease in greenhouse gases purify fossil-fuel energy sources; then, from the standpoint of
(GHGs) emissions. For instance, emissions targeting and planning process integration (Shenoy, 1995; El-Halwagi, 1997, 2006;
has been done by Atkins et al. (2008, 2009a, 2009b) for the New Smith, 2005; Kemp, 2007), CCS is analogous to the concept of
Zealand electricity industry, and by Crilly and Zhelev (2008, 2009) regeneration or interception in resource conservation networks
for the Irish electricity generation sector. such as those for water and hydrogen (Agrawal and Shenoy, 2006)
Carbon capture and storage (CCS), alternatively referred to as but with some differences. One basic difference is the possibility
carbon capture and sequestration, is a technology for mitigating of multiple CCS sources in carbon emission networks (CENs),
the contribution of carbon emissions to global warming, by which is in contrast to typically a single regeneration stream in
capturing carbon dioxide (CO2) from large point sources such as water and hydrogen networks. The problem of targeting multiple
fossil-fired power plants and storing it in a manner to prevent it resources has been studied earlier in the context of heat exchan-
from entering the atmosphere. Some CCS techniques (Steeneveldt ger networks (Shenoy et al., 1998; Shethna et al., 1999), mass
et al., 2006; Wall, 2007; Yang et al., 2008) with capability for exchanger networks (Fraser et al., 2005) and resource allocation
minimally 80% CO2 removal include post-combustion capture networks (Shenoy and Bandyopadhyay, 2007). Here, the unified
(PCC where CO2 is absorbed from flue gases using chemical conceptual approach (Agrawal and Shenoy, 2006; Prakash and
agents), oxyfuel combustion (Oxyf where pure oxygen rather Shenoy, 2005a), which essentially depends on the composite table
algorithm (CTA) for targeting and the nearest neighbors algorithm
(NNA) for network design, is adapted to handle multiple CCS
n
Corresponding author. Tel.: þ91 22 40104615. sources as demonstrated below. The unified conceptual approach
E-mail address: shenoys@vsnl.com (U.V. Shenoy). (with the CTA and NNA) has been successfully used by Deng and

0009-2509/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2011.09.041
314 A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327

Feng (2009) for an optimal water network with regeneration/ available. Each resource has a specified carbon intensity/emission
recycle in an alumina plant to obtain zero wastewater discharge factor CRk and a cost per energy unit cRk. The excess (or unused)
and significant freshwater savings. energy (denoted by subscript E) may be treated as an external
The proposed algorithmic procedure for CENs is associated demand or waste (Shenoy and Bandyopadhyay, 2007) without
with a graphical representation called the limiting composite any restrictions.
curve, in which net deficit/surplus may be conveniently repre- The network for the energy allocation problem with CCS is
sented on an emission factor versus emission load diagram represented in the form of a matching matrix (Prakash and
(Shenoy, 2010) similar to the temperature versus enthalpy Shenoy, 2005b; Das et al., 2009; Shenoy, 2010) in Fig. 1. The
diagram in heat exchanger network synthesis (Shenoy, 1995) or values of emission load M, energy F, and emission factor C are
the concentration-mass diagram in mass exchanger network non-negative real numbers. The three quantities are related by
synthesis (El-Halwagi and Manousiouthakis, 1990) or the water M ¼ FC ð1Þ
purity versus contaminant load diagram in the design of water
networks (Wang and Smith, 1994; Hallale and Fraser, 1998). The In general, the following conservation balance equations and
concepts and algorithms are illustrated below through a case linear mixing rules fundamentally hold for energy and emission
study based on the data given by Pekala et al. (2010), who studied load, when sources S1 and S2 are mixed to satisfy demand D.
CCS options to retrofit fossil-fuel power plants. F S1 þ F S2 ¼ F D ð2aÞ

F S1 C S1 þ F S2 C S2 ¼ F D C D ð2bÞ
2. Problem statement and mathematical formulation
The objective is to match sources (after appropriate mixing, if
necessary, as per Eqs. (2a,b)) and demands, so as to minimize the
Based on the problem statement for carbon-constrained
operating cost OC given by the following expression:
energy planning (Tan and Foo, 2007; Pekala et al., 2010), a
detailed mathematical formulation for the targeting problem of NR
X NS
X NS
X
energy allocation with carbon capture and storage (CCS) is OC ¼ cRk F Rk þ c0Si F 0Si þ cSi ðF Si f iE Þ ð3aÞ
k¼0 i¼1 i¼1
developed below. A set of ND demands (regions) and a set of NS
sources are given. Each demand Dj has an energy requirement FDj In Eq. (3a), the three terms signify clean-carbon resources, the
and a carbon footprint or emission load that must not exceed a sources with CCS, and the sources without CCS (excluding the
specified maximum limit MDj. Each source has a known avail- excess or unused energy fiE), respectively. On noting that the CCS
ability FSi,total. Without CCS, each source Si has a specified carbon and non-CCS portions of each source total the known availability
intensity/emission factor CSi and a cost per energy unit cSi. With (i.e., FSi0 þFSi ¼FSi,total), Eq. (3a) may be rewritten as
CCS (denoted by single prime), each source Si0 has a specified NR NS NS NS
X X X X
emission factor C 0Si and a cost per energy unit c0Si . A set of (NR þ1) OC ¼ cRk F Rk þ ðc0Si cSi ÞF 0Si þ cSi F Si,total  cSi f iE ð3bÞ
clean-carbon resources (zero-carbon R0 and/or low-carbon Rk) is k¼0 i¼1 i¼1 i¼1

Fig. 1. General representation of an energy allocation problem with CCS as a matching matrix.
A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327 315

In Eq. (3b), the four terms signify the cost of clean-carbon In an analogous manner, the overall system emission load
resources, the additional cost of sources with CCS (with reference balance (as per Eq (2b)) may be written as
to the non-CCS option), the cost of total available sources (which ND NS
X X
is a constant), and the cost of excess or unused energy, respec- F R C R F E C E ¼ D2 where D2  F Dj C Dj  ðF 0Si C 0Si þ F Si C Si Þ
tively. Since the third term is constant and the fourth term is zero j¼1 i¼1
for the case of CCS with no excess energy, the objective function ð7cÞ
to be minimized consists of the first two terms in Eq. (3b). For the
case of CCS with zero excess, Eq. (3b) importantly indicates that As per Eqs. (7b) and (7c), the net system energy deficit (D1)
and the net system emission load deficit (D2) are obtained by
CCS sources may be regarded as clean-carbon resources with their
cost per energy unit taken as (c0Si  cSi). subtracting the sum of all sources (CCS and non-CCS) from the
sum of all demands. Defining such net system quantities, which
Let fij denote the energy allocated from non-CCS source i to
0
demand j, and f ij denote the energy allocated from CCS source i to are constant for a given problem, is useful (Bandyopadhyay et al.,
1999, 2004; Shenoy, 2010) in process systems engineering.
demand j. Further, let fkj represent the energy allocated from
clean-carbon resource k to demand j, and fiE designate the excess On using the emission load conservation in Eq. (2b), the carbon
footprint or emission load limit for each demand may be
energy from non-CCS source i. Thus, f symbolizes the energy
allocations for the cell matches of the matching matrix in Fig. 1. expressed as
Due to the energy conservation in Eq. (2a), the energy balance for NR
X NS
X NS
X
0
each demand and each non-CCS source is given by f kj C Rk þ f ij C 0Si þ f ij C Si r F Dj C Dj
k¼0 i¼1 i¼1
NR
X NS
X 0
NS
X for every demand j A ND ð8aÞ
f kj þ f ij þ f ij ¼ F Dj for every demand j A ND ð4aÞ
k¼0 i¼1 i¼1 An alternative (Pekala et al., 2010) to Eq. (8a) would be an
aggregated emission limit (i.e., total emissions of all demands
ND
must not be exceeded), which is obtained by summing over all
X
f ij þ f iE ¼ F Si for every non-CCS source i A NS ð4bÞ demands. Further, if the availability of clean-carbon resources is
j¼1 restricted, then each resource may be limited in general to a
specified maximum FRk,max as
The total requirement of each CCS source and each resource is
F Rk rF Rk,max for every resource k A NR þ 1 ð8bÞ
calculated from
The objective is to minimize the operating cost OC in Eq. (3)
ND
X 0
f ij ¼ F 0Si for every CCS source iA NS ð5aÞ subject to the constraints given by Eqs. (4), (5), and (8). As the
j¼1 objective function and all the constraints are linear, this is a linear
programming (LP) formulation.
Besides, note that the problem formulation for targeting
ND
X multiple resources (Shenoy and Bandyopadhyay, 2007) becomes
f kj ¼ F Rk for every resource k A NR þ 1 ð5bÞ
j¼1
a special case of the above formulation wherein CCS sources are
not present.
Eq. (4a) corresponds to the column sum, whereas Eqs (4b), The above LP formulation is solved below through an algebraic
(5a), and (5b) correspond to the row sum for the energy values in procedure and validated using optimization software. The pro-
the matching matrix (Fig. 1). As mentioned earlier, the CCS and posed algebraic procedure is based on the limiting composite
non-CCS amounts of each source total the specified availability curve for carbon emission networks (Shenoy, 2010), whose data
according to are readily obtained through the Composite Table Algorithm
(CTA), first proposed by Agrawal and Shenoy (2006).
F 0Si þF Si ¼ F Si,total for every source i A NS ð5cÞ

The total excess energy and the total resource requirement 3. Representation of CCS sources
(using Eq. (5b)) may be expressed as

NS
X The starting point for the Composite Table Algorithm (CTA) is
FE ¼ f iE ð6aÞ the representation of sources (and demands) by vertical arrows to
i¼1 conveniently determine their populations in the different emis-
sion-factor intervals. For each source (and demand), the arrow
NR
X NR X
X ND begins at its specified emission factor. All arrows end at an
FR ¼ F Rk ¼ f kj ð6bÞ arbitrarily chosen large value of the emission factor. Fig. 2a shows
k¼0 k¼0j¼1 the representation used in the CTA in terms of separate arrows for
a CCS source (with emission factor CSi0 and energy availability FSi0 )
On taking the summation over all demands and all non-CCS and its corresponding non-CCS source (with emission factor CSi
sources (using Eqs. (4a) and (4b), respectively), the following and energy availability FSi).
overall energy balance over the system is established. Fig. 2b shows an equivalent composite representation
NR X
ND NS X
ND NS ND NS obtained by appropriately adding the energy values in the two
X X X X X
f kj þ
0
f ij þ F Si ¼ F Dj þ f iE ð7aÞ emission-factor intervals. Combining gives the energy value as F 0Si
k¼0j¼1 i¼1j¼1 i¼1 j¼1 i¼1 for C 0Si oCrCSi, and the energy value as F 0Si þFSi ¼FSi,total for C 4CSi.
Thus, in the equivalent composite representation, the non-CCS
Using Eqs. (5a),(6a) and (6b), the above overall system energy source is represented by an arrow with a known constant energy
balance may be compactly rewritten as value of FSi,total (rather than an unknown variable FSi). Further, in
ND NS
the equivalent composite representation, the CCS source is
X X
F R F E ¼ D1 where D1  F Dj  ðF 0Si þ F Si Þ ð7bÞ represented by a line-segment between C 0Si and CSi (rather than
j¼1 i¼1 an arrow from CSi0 to an arbitrary large emission-factor value).
316 A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327

C’Si C’Si

F’Si F’Si

CSi CSi

F’Si FSi FSi,total

CCS Source Si’ non-CCS Source Si CCS Source Si’ non-CCS Source Si

Fig. 2. Arrow representation of CCS source and corresponding non-CCS source in CTA (a) separately and (b) as equivalent composite.

The equivalent composite representation has two important Table 1


implications. First, the implementation of the CTA from Shenoy Data for case study.
(2010) remains unchanged requiring the demands and only the
Demands/sources Emission factor C Energy F Emission M Cost factor c
non-CCS sources (with energy values of FSi,total) to be considered.
Second, the CCS sources (represented as line segments) can be Demand data
regarded as resources and targeted by profile matching with the D1 (region I) (20) 100 20
limiting composite curve. Note that the limiting composite for D2 (region II) (50) 40 20
D3 (region III) (100) 60 60
carbon emission networks developed by Shenoy (2010) is thus Total 200 100
sacrosanct: it remains unchanged for targeting with and without
Source data
CCS because it itself does not include the CCS sources. Further-
R0 (zero-C resource) 0 0.045
more, the number of targeting variables associated with sources is S10 (CCS—natural gas) 10 0.042
halved because only CCS sources are targeted whereas non-CCS S20 (CCS—oil) 15 0.034
sources (with their total energy availability) are included in the S30 (CCS—coal) 20 0.032
CTA. Both the CTA implementation and the profile matching are S1 (natural gas) 55 20 (11) 0.025
S2 (oil) 75 80 (60) 0.022
illustrated in the next section through a case study. S3 (coal) 105 60 (63) 0.020
Total 160 134

Units: emission factor C in t CO2/TJ, energy F in 104 TJ, emission load M in


106 t CO2, cost factor c in 1/TJ.

4. Targeting by profile matching with limiting composite


4.1. Clean-carbon resource with No CCS source
To illustrate the methodology, the case study from Pekala et al.
(2010), which is itself adapted from Tan and Foo (2007), is For the case of zero-carbon (or completely CO2-neutral)
considered. The data for the three demands (three regions with resource (i.e., CR ¼CR0 ¼0), referred to as Case A henceforth,
their expected energy consumptions FDj in TJ and emission load targeting may be graphically done by rotating the horizontal axis
limits MDj in t CO2) and the three sources (natural gas, oil and coal with the origin as pivot until it just touches the limiting
with their emission factors CSi in t CO2/TJ and total energy composite curve. The reciprocal of the slope of this rotated
availabilities FSi,total in TJ) are given in Table 1. The values for CDj resource target line gives the minimum zero-carbon resource
in t CO2/TJ and MSi in t CO2 (in parenthesis) in Table 1 are readily requirement. The point where the target profile touches the
calculated using Eq. (1). The emission factors C 0Si in t CO2/TJ for the limiting composite defines the pinch (Fig. 3). This graphical
three sources with CCS (S10 , S20 and S30 ) are also provided. The procedure (Shenoy, 2010) is identical in principle to that sug-
only clean-carbon resource is a zero-carbon resource (R0). Oper- gested by Wang and Smith (1994) for the determination of the
ating cost factors for all sources (zero-carbon resource, three CCS minimum freshwater target.
sources, and three non-CCS sources) are given in unspecified cost The mathematical alternative is to use the following equation
units (Pekala et al., 2010) per TJ of energy (relative to baseline based on an emission load balance over the below-pinch region:
power generation costs).
MP ¼ F R ðC P 2C R Þ ð9Þ
The implementation of the CTA (with details available in Shenoy,
2010) for this case study is shown in Table 2. The data in Table 2 where MP is the cumulative emission load of the pinch point on the
may also be readily generated by the Unified Targeting Algorithm limiting composite, CP is the pinch emission factor, and FR is the
(UTA), which is a generalized CTA recently proposed by Shenoy energy requirement target for resource R with emission factor CR.
(2011). The emission-factor C (in the first column) is plotted against Since the pinch point is not known a priori, the following
the cumulative emission load M (in the fourth column) to obtain the procedure is adopted (Agrawal and Shenoy, 2006; Shenoy, 2010).
limiting composite curve (dashed line in Fig. 3). The limiting In accordance with Eq. (9), the possible amount of resource
composite provides the net deficit (as a function of emission factor) required is calculated on dividing the cumulative emission load
after considering all demands and non-CCS sources with their total M by (C  CR) in Table 2 and tabulated in the last column. The
energy availabilities FSi,total. The clean resource(s) and the CCS possible resource requirement corresponds to the reciprocal of
sources are excluded because the aim is to target their minimum the slope of a line originating from CR on the vertical axis to a
requirements as demonstrated next. point on the limiting composite curve. Since the resource line can
A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327 317

Table 2
Limiting composite curve by Composite Table Algorithm (CTA).

150 zero-carbon resource requirement (with FE ¼0, i.e., zero excess) as


FR ¼FR0 ¼400,000 TJ.
125 For the case of clean-carbon resource with no CCS sources, the
target profile is a single straight line of constant slope (as in Fig. 3)
Emission Factor (tCO2/TJ)

100 starting from CR0. When a single CCS source is used along with the
Limiting Composite
Pinch clean-carbon resource, the target profile is the composite of two
75
lines: one for the clean-carbon resource starting from CR0, and
another for the CCS source starting from CSi0 and terminating at
CSi. In general, the target profile will comprise three sections,
50
Target Profile whose arrow representation is shown in Fig. 4a and whose
equations based on emission load balances are as follows:
25 1/FR0
M ¼ F R0 ðC2C R0 Þ for C r C 0Si ð10aÞ
0
0 10 20 30 40 50 60 70 80 90
M ¼ F R0 ðC2C R0 Þ þ F 0Si ðC2C 0Si Þ for C 0Si r C r C Si ð10bÞ
Emission Load or Carbon Footprint (106 t CO2)

Fig. 3. Target profile for zero-carbon resource with no CCS source. (Case A with M ¼ F R0 ðC2C R0 Þ þ F 0Si ðC Si 2C 0Si Þ for C Z C Si ð10cÞ
FR0 ¼ 813,333 TJ).
As before, the target profile must just touch the limiting
composite curve (at the pinch) and must always be below it.
never be above the limiting composite, the minimum slope is Since the pinch is not known a priori, a procedure similar to
required; so, the maximum value in the last column of Table 2 that adopted earlier with Eq. (9) is used. With FR0 ¼400,000 TJ
gives the minimum resource target. Thus, the zero-carbon (the absolute minimum zero-carbon resource requirement in this
resource target FR is 813,333 TJ (as per Table 2) and the corre- case), F 0Si is the only unknown variable and may be readily
sponding emission factor specifies the pinch (75 t CO2/TJ). calculated from Eqs. (10b) or (10c) for the different points
From Eqs. (7b) and (7c), D1 ¼(200–160)  104 ¼400,000 TJ and (M, C) of the limiting composite as shown in Table 3. With CR0 ¼0
D2 ¼(100–134)  106 ¼ 34  106 t CO2 on using Table 1 and not (for zero-carbon resource), substitution in Eqs. (10b) and (10c)
considering CCS sources. Since FR ¼813,333 TJ and CR ¼0, Eqs. (7b) yields the values in Table 3 for F 0S1 (third column with C 0S1 ¼10 and
and (7c) yield FE ¼413,333 TJ and CE ¼82.258 t CO2/TJ (which is CS1 ¼55), F 0S2 (fourth column with C 0S2 ¼15 and CS2 ¼75), and F 0S3
the minimum emission factor possible for the excess energy when (fifth column with C 0S3 ¼20 and CS3 ¼105). As before, the highest
demands are met at the maximum emission load—the limit of value in each of these three columns gives the target as
maximum environmental cost and minimum economic/operating F 0S1 ¼755,556 TJ (with pinch at 105 t CO2/TJ), F 0S2 ¼566,667 TJ
cost). In essence, the minimum targets for zero-carbon resource (with pinch at 105 t CO2/TJ) and F 0S3 ¼563,636 TJ (with pinch at
and excess energy are 813,333 and 413,333 TJ, respectively. These 75 t CO2/TJ). For the target to be feasible, it must not be greater
targets and the carbon emission networks that satisfy the targets than the total available FSi,total. Although the target is feasible for
are discussed in detail by Shenoy (2010). S2 (566,667 TJ o800,000 TJ) and S3 (563,636 TJo600,000 TJ), it is
not for S1 (755,556 TJ 4200,000 TJ). Since it is not possible to use
4.2. Single CCS source the absolute minimum zero-carbon resource (FR0 ¼400,000 TJ),
the target for S1 is reworked. On setting F 0S1 ¼200,000 TJ (max-
The zero-carbon resource requirement can be reduced by imum possible based on availability), Eqs. (10b) and (10c) yield
using a less expensive CCS source. The absolute minimum zero- the values in the last column of Table 3 for FR0 with the highest
carbon resource can be targeted in a manner similar to that used value providing the target as FR0 ¼693,333 TJ (with pinch at
for regeneration with recycle for water networks (Wang and 75 t CO2/TJ).
Smith, 1994; Agrawal and Shenoy, 2006; Deng and Feng, 2009) Target profiles using Eqs. (10a)–(10c) are plotted in Fig. 5 for
as illustrated next. The slope of the limiting composite below C 0Si each one of the three CCS sources (S10 , S20 , and S30 ) along with the
sets the minimum FR or FE target in accordance with Eq. (7b). For zero-carbon resource R0. The pinch, where the target profile
this case study, the limiting composite starts at 20 t CO2/TJ touches the limiting composite, is expectedly observed at
(Table 2) and the values of C 0Si are 10, 15 and 20 t CO2/TJ 105 t CO2/TJ (in Fig. 5b for S20 , referred to as Case B2 hereafter)
(Table 1). Therefore, the slope of the limiting composite below and at 75 t CO2/TJ (in Fig. 5c for S30 (Case B3) and in Fig. 5d for the
CSi0 is infinite, and its reciprocal sets the minimum FR or FE as reworked feasible case of S10 (Case B1)). The infeasible case of S10
zero. Since D1 ¼400,000 TJ, Eq. (7b) gives the absolute minimum (with F 0S1 ¼ 755,5564200,000 TJ) is shown in Fig. 5a.
318 A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327

FR0 FR0
CR0 CR0
Eq. (10a) Eq. (11a)
F’Si F’Si
C’Si C’Si
Eq. (11b)
F’S(i+1)
Eq. (10b) C’S(i+1)
Eq. (11c)
CSi CSi
Eq. (11d)
Eq. (10c) CS(i+1)
Eq. (11e)
R0 Si’ R0 Si’ S(i+1)’

Fig. 4. Arrow representation of target profile for clean-carbon resource R0 with (a) single CCS source and (b) two CCS sources.

Table 3
Targeting for single CCS source.

Units: emission factor C in t CO2/TJ, energy F in 104 TJ, emission load M in 106 t CO2.

4.3. Two CCS sources M ¼ F R0 ðC2C R0 Þ þ F 0Si ðC Si 2C 0Si Þ þ F 0Sði þ 1Þ ðC2C 0Sði þ 1Þ Þ

for C Si r C rC Sði þ 1Þ ð11dÞ


On considering two CCS sources along with the zero-carbon
resource R0, three cases exist: Case C1 with (R0, S10 , S20 ); Case C2
with (R0, S10 , S30 ); and Case C3 with (R0, S20 , S30 ). Targeting for M ¼ F R0 ðC2C R0 Þ þ F 0Si ðC Si 2C 0Si Þ þ F 0Sði þ 1Þ ðC Sði þ 1Þ 2C 0Sði þ 1Þ Þ
each of these cases is discussed next. for C ZC Sði þ 1Þ ð11eÞ
When two CCS sources are used along with the clean-carbon
resource, the target profile is the composite of three lines: one for With FR0 ¼400,000 TJ (the absolute minimum zero-carbon
the clean-carbon resource R0 starting from CR0, second for the CCS resource requirement) and F 0Si ¼F 0S1 ¼ 200,000 TJ (the maximum
source Si starting from C 0Si and terminating at CSi, and third for the availability of S10 ), FS(i
0
þ 1) is the only unknown variable and may
CCS source S(iþ1) starting from CS(i þ 1)0 and terminating at CS(i þ 1). be easily calculated from Eqs. (11c),(11d) or (11e) for the different
Here, the notation i and (iþ1) is used for convenience to denote points (M, C) of the limiting composite in Table 4. With CR0 ¼0 (for
any two sources, which may not actually be consecutive. In zero-carbon resource), C 0Si ¼C 0S1 ¼10 and CSi ¼CS1 ¼55, substitution
general, the target profile will comprise five sections, whose in Eqs. (11c),(11d) and (11e) gives the values in Table 4 for F 0S2
arrow representation is shown in Fig. 4b and whose equations (third column with C 0S2 ¼15 and CS2 ¼75), and F 0S3 (fourth column
based on emission load balances are as follows: with C 0S3 ¼20 and CS3 ¼105). The highest value in each column
gives the target as F 0S2 ¼416,667 TJ (with pinch at 105 t CO2/TJ)
M ¼ F R0 ðC2C R0 Þ for C r C 0Si ð11aÞ
and F 0S3 ¼400,000 TJ (with pinch at 75 t CO2/TJ). The targets, being
less than the total available FSi,total, are feasible for S2
M ¼ F R0 ðC2C R0 Þ þF 0Si ðC2C 0Si Þ for C 0Si r C rC 0Sði þ 1Þ ð11bÞ (416,667o800,000) as well as S3 (400,000o600,000). Profiles,
based on these targets and Eqs. (11a)–(11e), are plotted in Fig. 6a
and b for Case C1 with (R0, S10 , S20 ) and Case C2 with (R0, S10 , S30 ),
M ¼ F R0 ðC2C R0 Þ þF 0Si ðC2C 0Si Þ þ F 0Sði þ 1Þ ðC2C 0Sði þ 1Þ Þ respectively. The pinch, as expected, is seen at 105 t CO2/TJ (in
for C 0Sði þ 1Þ r C rC Si ð11cÞ Fig. 6a) and at 75 t CO2/TJ (in Fig. 6b).
A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327 319

150 150

125 Pinch 125


Emission Factor (tCO2/TJ)

Emission Factor (tCO2/TJ)


100 100 Limiting Composite
Limiting Composite
Pinch
75 1/FR0 75
Target Profile
50 50
Target Profile
1/(FR0 +F’S3)
25 1/(FR0 +F’S1) 25
1/FR0
1/FR0
0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Emission Load or Carbon Footprint (106 t CO2) Emission Load or Carbon Footprint (106 t CO2)

150 150

125 125
Pinch

Emission Factor (tCO2/TJ)


Emission Factor (t O2/TJ)

Limiting Composite
100 100
Limiting Composite 1/FR0
Pinch
75 75
1/FR0
50 Target Profile 50
Target Profile
1/(FR0 +F’S2)
25 25 1/(FR0 + F’S1)

1/FR0 1/FR0
0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
Emission Load or Carbon Footprint (106 t CO2) Emission Load or Carbon Footprint (106 t CO2)

Fig. 5. Target profile for single CCS source (a) infeasible case with F 0S1 ¼ 755556 TJ and FR0 ¼400,000 TJ, (b) Case B2 with F 0S2 ¼566,667 TJ and FR0 ¼ 400,000 TJ, (c) Case B3
with FS30 ¼563,636 TJ and FR0 ¼ 400,000 TJ, and (d) Case B1 with F 0S1 ¼ 200,000 TJ and FR0 ¼693,333 TJ.

Table 4 (F 0Si and FS(iþ


0
1) ). Two unknown variables imply the existence of two
Targeting for Two CCS Sources. pinch points. The approach thus far suggests the simultaneous
solution of two equations from Eqs. (11b)–(11e) for every possible
pair of points (M, C) on the limiting composite. However, a simpler
approach fruitfully utilizes the fact that all points on the limiting
composite are not candidate pinch points. A point on the limiting
composite is a candidate pinch only if the slope of the curve increases
after the point and consequently the point ‘protrudes toward’ the
target profile. Thus, the net energy deficit in an interval (Fnet as given
by the second column in Table 2), which corresponds to the reciprocal
of the slope of a segment on the limiting composite, must decrease at
candidate pinch points. In this case study, the candidate pinches are
at emission factors of 55 t CO2/TJ (where Fnet decreases from 140 to
120 TJ), 75 t CO2/TJ (where Fnet decreases from 120 to 40 TJ), and
105 t CO2/TJ (where Fnet decreases from 100 to 40 TJ). However, the
pinch at 55 t CO2/TJ is possibly not a candidate (because of the
relatively small change in the Fnet value and the slope/shape of the
limiting composite curve).
With the two pinch points at 75 t CO2/TJ and 105 t CO2/TJ (as
observed in all the six cases analyzed so far), Eqs. (11c) and (11d) give

61  106 ¼ F R0 ð7520Þ þ F 0S2 ð75215Þ þ F 0S3 ð75220Þ ð12aÞ


4 6
Units: emission factor C in t CO2/TJ, energy F in 10 TJ, emission load M in 10 t CO2.
76  106 ¼ F R0 ð10520Þ þ F 0S2 ð75215Þ þF 0S3 ð105220Þ ð12bÞ
0 0
The third case of (R0, S2 , S3 ), referred to as Case C3, is now Subtracting Eq. (12a) from Eq. (12b) yields
explored. With FR0 ¼400,000 TJ (the absolute minimum zero-carbon
resource requirement), Eqs. (11b)–(11e) have two unknown variables ð76261Þ  106 ¼ ðF R0 þ F 0S3 Þð105275Þ ð12cÞ
320 A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327

150
Emission Factor (t CO2/TJ)

125 Pinch

100
Limiting Composite 1/FR0

75
1/(FR0 + F’S2)
50 Target Profile 150
1/(FR0 + F’S1+F’S2)
25

Emission Factor (t CO2/TJ)


125
1/(FR0 + F’S1) Pinches
1/FR0
0 1/FR0
0 10 20 30 40 50 60 70 80 90 100
Emission Load or Carbon Footprint (106t CO2) 1/(FR0 + F’S3)
75
Limiting Composite
150 50 1/(FR0 + F’S2+F’S3)
Target Profile
Emission Factor (t CO2/TJ)

125 25
1/(FR0 + F’S2)
100 Limiting Composite 0
1/FR0

Pinch 1/FR0 0 10 20 30 40 50 60 70 80 90
75 Emission Load or Carbon Footprint (106t CO2)
1/(FR0 + F’S3)
50
Target Profile

25 1/(FR0 + F’S1+F’S3)
1/(FR0 + F’S1)
1/FR0
0
0 10 20 30 40 50 60 70 80 90
Emission Load or Carbon Footprint (106t CO2)

Fig. 6. Target profile with FR0 ¼400,000 TJ for two CCS sources (a) Case C1 with F 0S1 ¼ 200,000 TJ and F 0S2 ¼416,667 TJ, (b) Case C2 with F 0S1 ¼200,000 TJ, and F 0S3 ¼400,000 TJ
(c) Case C3 with F 0S2 ¼ 425,000 TJ and F 0S3 ¼ 100,000 TJ.

Eq. (12c) is simply an emission load balance between the two source S(iþ1) starting from C 0Sði þ 1Þ and terminating at CS(i þ 1), and
pinches (whereas Eq. (12a) is a balance below the pinch at fourth for the CCS source S(i þ2) starting from C 0Sði þ 2Þ and
75 t CO2/TJ, and Eq. (12b) is a balance below the pinch at terminating at CS(i þ 2). In general, the target profile will comprise
105 t CO2/TJ) and provides a direct way to calculate the F 0S3 target. seven sections [C rC 0Si ; C 0Si rCrC 0Sði þ 1Þ ; C 0Sði þ 1Þ rC rC 0Sði þ 2Þ ;
If CSi and CS(i þ 1) denote the two pinch points (with the corre- C 0Sði þ 2Þ rC rCSi; CSi rC rCS(i þ 1); CS(i þ 1) rC rCS(i þ 2); and
sponding emission loads on the limiting composite given by MSi CZCS(i þ 2)], whose arrow representation can be shown in a form
and MS(i þ 1)), then Eq. (12c) in generalized form may be written as similar to Fig. 4 (omitted for brevity) and whose emission load
F R0 þ F 0Sði þ 1Þ ¼ ðM Sði þ 1Þ 2MSi Þ=ðC Sði þ 1Þ 2C Si Þ ð12dÞ balance equations can be written in a manner similar to Eqs. (10)
and (11) (not shown).
Equations (similar to Eq. (12d)) for the emission load balance With FR0 ¼400,000 TJ (the absolute minimum zero-carbon
between two pinches can be written for any of the sections in Fig. 4 resource requirement) and F 0Si ¼F 0S1 ¼ 200,000 TJ (the maximum
and readily used to establish targets in some multi-pinch cases. availability of S10 ), there remain two unknown variables (F 0S2 and
For Case C3 with (R0, S20 , S30 ), Eq. (12c) gives the target as F 0S3 ). As before, this suggests the existence of two pinch points (at
F 0S3 ¼100,000 TJ and then Eq. (12a) or (12b) give F 0S2 ¼425,000 TJ 75 and 105 t CO2/TJ). An emission load balance between the two
on substituting FR0 ¼400,000 TJ. The targets, being less than the pinches yields Eq. (12c) and the target as F 0S3 ¼100,000 TJ (same as
total available FSi,total, are feasible for both S2 (425,000o800,000) in the earlier case). Now, balances below the pinches at 75 t CO2/
and S3 (100,000o600,000). The profile for Case C3 with (R0, S20 , TJ and 105 t CO2/TJ give the following equations (which are the
S30 ), based on these targets and Eqs. (11a)–(11e), is plotted in analogs of Eqs. (12a) and (12b)):
Fig. 6c and expectedly shows two pinches (at 75 and 105 t CO2/TJ).
61  106 ¼ F R0 ð7520Þ þ F 0S1 ð55210Þ þ F 0S2 ð75215Þ þ F 0S3 ð75220Þ
4.4. Three CCS sources ð13aÞ

The last case (Case D) considers all three CCS sources along 76  106 ¼ F R0 ð10520Þ þ F 0S1 ð55210Þ þ F 0S2 ð75215Þ þ F 0S3 ð105220Þ
with the zero-carbon resource R0.
When three CCS sources are used along with the clean-carbon ð13bÞ
resource, the target profile is the composite of four lines: one for
the clean-carbon resource R0 starting from CR0, second for the CCS On substituting FR0 ¼400,000 TJ, F 0S1 ¼200,000 TJ and F 0S3 ¼
source Si starting from C 0Si and terminating at CSi, third for the CCS 100,000 TJ, the target is obtained from Eq. (13a) or (13b) as
A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327 321

150

125
Pinches

Emission Factor (t CO2/TJ)


1/FR0
100

1/(FR0 + F’S3)
75
Limiting Composite
1/(FR0 + F’S2 + F’S3)
50
Target Profile
25 1/(FR0 + F’S1 + F’S2 + F’S3)
1/(FR0 + F’S1 + F’S2)
1/(FR0 + F’S1)
1/FR0
0
0 10 20 30 40 50 60 70 80 90
Emission Load or Carbon Footprint (106t CO2)

Fig. 7. Target profile with FR0 ¼ 400,000 TJ for three CCS sources: Case D with F 0S1 ¼200,000 TJ, F 0S2 ¼ 275,000 TJ and F 0S3 ¼ 100,000 TJ.

Table 5
Summary of results for case study.

No CCS Single CCS Two CCS Three CCS

S10 S20 S30 (S10 , S20 ) (S10 , S30 ) (S20 , S30 ) (S10 , S20 , S30 )
Case A Case B1 Case B2 Case B3 Case C1 Case C2 Case C3 Case D

FR0 (104 TJ) 81.3333 69.3333 40 40 40 40 40 40


F 0S1 (104 TJ) – 20 – – 20 20 – 20
F 0S2 (104 TJ) – – 56.6667 – 41.6667 – 42.5 27.5
F 0S3 (104 TJ) – – – 56.3636 – 40 10 10
Pinch CP (t CO2/TJ) 75 75 105 75 105 75 75, 105 75, 105
Limiting composite Fig. 3 Fig. 5d Fig. 5b Fig. 5c Fig. 6a Fig. 6b Fig. 6c Fig. 7
Operating cost target – – 59,400 59,364 61,000 60,800 58,900 60,500
Networks by NNA Shenoy (2010) Fig. 8a Fig. 8b Fig. 8c Fig. 9a Fig. 9b Fig. 9c Fig. 10
Operating cost 62,307 62,947 59,400 59,364 61,000 60,800 58,900 60,500
Fixed cost 0 1000 800 700 1800 1700 1500 2500
Total cost 62,307 63,947 60,200 60,064 62,800 62,500 60,400 63,000

F 0S2 ¼275,000 TJ. The targets, being less than the total availability, the sources to be chosen are the nearest available neighbors to
are feasible for S2 (275,000o800000) and S3 (100,000o the demand in terms of emission factor.’ Details of the NNA are
600,000). The profile for Case D with (R0, S10 , S20 , S30 ) based on available in these works and are not discussed here.
these targets is plotted in Fig. 7 and expectedly shows two The NNA, which is a network design tool and should be applied
pinches (at 75 and 105 t CO2/TJ) as well as seven sections. only after targets are established, is conveniently implemented in
The targets for all eight primary cases (only R0 with no CCS, terms of the matching matrix representation (Prakash and Shenoy,
three cases with single CCS source, three cases with two CCS 2005b; Das et al., 2009; Shenoy, 2010). All CENs (Figs. 8, 9, and 10)
sources, and a final case with all three CCS sources) are summar- in this work are presented in the form of matching matrices.
ized in Table 5 (first six rows). The target for the operating cost A matching matrix provides a superstructure with a placeholder
(seventh row) as per Eq. (3) can be established a priori in six cases for every possible match in the allocation network. The emission
(where the excess energy is zero, i.e., when FR0 ¼400,000 TJ) with loads or carbon footprints [shown in square brackets in the
the exception of Case A and Case B1. bottom half of each relevant cell of the matching matrix for ready
reference] are calculated by simple multiplication using Eq. (1)
only after completing the design. These values are not required to
5. Network design by Nearest Neighbors Algorithm (NNA) be calculated during the actual design of the network by the NNA.
In the matching matrix, it is convenient to arrange sources in order
In this section, networks are synthesized to meet the targets of increasing (or decreasing) emission factor to promptly identify
established. The Nearest Neighbors Algorithm (NNA), proposed by the nearest neighbors; however, it is not essential (although often
Prakash and Shenoy (2005a) for water networks and discussed by convenient) to arrange demands because they can be met in any
Agrwal and Shenoy (2006) for water/hydrogen networks as well order. Arranging demands in order of emission factor nonetheless
as by Shenoy (2010) for carbon emission networks (CENs), is used allows easy identification of the cross-pinch regions (gray cells in
for this purpose. In the context of CENs, the basic principle behind the matching matrices that follow). To achieve the targets using
the NNA may be stated as follows: ‘To satisfy a particular demand, the NNA, there should be no cross-pinch energy transfer, which is
322 A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327

Fig. 8. Carbon emission networks with single CCS source using (a) S10 (Case B1), (b) S20 (Case B2), and (c) S30 (Case B3).

ensured by designing separately in the non-grayed below-pinch three networks using two CCS sources are shown in Fig. 9a (for
and above-pinch regions. Case C1), Fig. 9b (for Case C2) and Fig. 9c (for Case C3). Finally,
Many different networks, all satisfying the targets, may in a network using all three CCS sources is shown in Fig. 10 for Case
general be designed by the NNA depending on the order in which D. Of these seven networks, two networks (Figs. 8b and 9c) match
the demands are met (Agrawal and Shenoy, 2006; Shenoy, 2010). those reported by Pekala et al. (2010), who obtained them by
Here, one promising network synthesized by the NNA is solving a mixed integer linear programming (MILP) formulation
presented for each primary case taking into consideration that using optimization software and varying the fixed cost factors for
the number of CCS matches must be minimized to keep the fixed CCS sources. The network for Case A of only zero-carbon resource
cost low. In all the networks below, the NNA is applied to satisfy R0 (with no CCS sources) is not presented here because it has
the demands D1, D2 and D3 in that order. Three possible net- been dealt with in detail by Shenoy (2010).
works using only a single CCS source are shown in Fig. 8a (for In all the networks except those in Figs. 8c and 9b, the demands
Case B1), Fig. 8b (for Case B2) and Fig. 8c (for Case B3). Another are met at the maximum emission load limits or equivalently
A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327 323

Fig. 10. Carbon emission networks with three CCS sources using S10 , S20 , and S30
(Case D).

6. Cost analysis

The costs for the networks synthesized in the previous section


may now be evaluated as shown in Table 5 (last three rows). The
operating costs (third last row in Table 5) exactly match with
their respective targets (seventh row). In addition to the operating
cost factors given in the last column of Table 1, Pekala et al.
(2010) have specified the fixed cost factors (in unspecified cost
units) for matches using the three CCS sources (1000 for
CCS—Natural Gas match; 800 for CCS—Oil match; and 700 for
CCS—Coal match). Using these factors (of 1000, 800, and 700), the
fixed cost (second last row in Table 5) can be calculated and
added to the respective operating cost to obtain the total cost (last
row). Note that the operating costs as well as the total costs for
the various cases are within about 7% of the minimum. Further,
the operating costs are expectedly higher for cases using S10
(because of the relatively high operating cost factor of 0.042 for
S10 , being close to 0.045 for R0).
From Table 5, the network with the minimum operating cost
(58,900) is found to use one CCS—Oil match and one CCS—Coal
match (Fig. 9c). It has a total cost of 60,400, which is reported to
be the globally optimal total cost solution by Pekala et al. (2010)
based on an MILP formulation. However, this is not consistent
with Table 5, which shows the network using only one CCS—Coal
match (Fig. 8c) to have the minimum total cost (60,064). The
network in Fig. 8c, in addition to being optimal in terms of total
cost, satisfies demand D3 with a carbon footprint of just
46.091  106 t CO2 (which is below its specified maximum limit
of 60  106 t CO2). Thus, the network in Fig. 8c provides a superior
Fig. 9. Carbon emission networks with two CCS sources using (a) S10 , S20 (Case C1)
solution to that reported by Pekala et al. (2010) (Fig. 9c) because it
(b) S10 , S30 (Case C2), and (c) S20 , S30 (Case C3).
has a lower total cost (60,064 instead of 60,400), a lower carbon
footprint (46.091  106 rather than 60  106 t CO2 for demand D3)
maximum emission factors. In Figs. 8c and 9b, demand D3 (which is and one fewer match (8 instead of 9 matches).
the only demand above the pinch at 75 t CO2/TJ in this case) is In Table 5, the targeting results (in the first six rows) include
satisfied with a carbon footprint below its specified maximum limit eight primary cases (in the columns) because there are three CCS
of 60  106 t CO2. This is consistent with an important observation sources in this case study. In general, 2NS primary cases are expected
in this regard made by Prakash and Shenoy (2005a) for water if there are NS CCS sources. Note that the primary cases have been
networks and Shenoy (2010) for carbon emission networks. In the fundamentally established by profile matching with the limiting
context of CENs, the observation may be stated as: ‘to meet the composite curve, without using any specific cost data. Generally,
minimum resource target, all demands below the pinch must be when actual cost data are unavailable or uncertain, it may be
satisfied at emission factors equal to their maximum allowable assumed that the unit operating cost (i.e., the operating cost per
values, whereas demands above the pinch can be satisfied at energy unit) reduces with decrease in quality (i.e., increase in
emission factors less than their maximum allowable values’. carbon intensity/emission factor) of the resource or source. In what
324 A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327

follows, a better approach is presented wherein a cost criterion is be elegantly expressed as


established for the applicability of the (eight) primary cases.
DcSi o RC DcSði þ 1Þ where RC  ðC Ei 2C 0Si Þ=ðC Eði þ 1Þ 2C 0Sði þ 1Þ Þ ð15aÞ
Consider for concreteness the introduction of S10 as a partial
replacement for S20 (specifically, for the optimal solution to Here, CE is appropriately taken as the emission factor of either
change from Case C3 to Case D). Given the relatively high the corresponding non-CCS source or the pinch, whichever is
operating cost factor (0.042) of S10 (in Table 1), this is currently lower in value. Note that the line segments (see Table 4) for both
not optimal. However, a condition in terms of the operating cost S10 and S20 terminate at or before the pinches (at 75 and
factor may be derived for the use of S10 to be optimal by using 105 t CO2/TJ) and therefore CE is taken as the emission factor of
Eq. (13) as a starting point. On substituting FR0 ¼400,000 TJ and the corresponding non-CCS source (55 for S1 and 75 for S2) in
F 0S3 ¼100,000 TJ, Eqs. (13a) and (13b) both give F 0S1 (55– Eq. (14d), which may not always be case. As discussed earlier, the
10)þF 0S2 (75–15)¼25.5  106. This equation includes Case C3 optimal solution does not use any Si0 for a positive deviation (i.e.,
(F 0S2 ¼425,000 if F 0S1 ¼0) as well as Case D (F 0S2 ¼275,000 if DcSi ¼RCDcS(i þ 1) þ d), whereas it is does use Si0 (in place of S(iþ1)0 )
F 0S1 ¼200,000). On substituting for F 0S2 from Eqs. (13a) and (13b) for a negative deviation (i.e., DcSi ¼RCDcS(i þ 1)  d), where d is a
into Eq. (3b), the operating cost may be expressed as small deviation of DcSi from RC DcS(i þ 1).
NS
The RC cost criterion in Eq. (15a) is validated by solving the linear
X
OC ¼ DcSi F 0Si þK 1 ð14aÞ program (LP) formulated in Section 2 for 28 runs (considering both
i¼1 positive and negative deviations of the 14 rows tabulated in Table 6).
The LP runs may be taken using Eq. (3) with base cost factors as
OC ¼ DcS1 F 0S1 þ DcS2 F 0S2 þ K 2 ð14bÞ given in the last column of Table 1, i.e., cR0 ¼0.045, c0S1 ¼0.042,
c0S2 ¼ 0.034, c0S3 ¼0.032, cS1 ¼0.025, cS2 ¼0.022, and cS3 ¼0.020. An
OC ¼ ½DcS1 DcS2 ð5510Þ=ð7515ÞF 0S1 þK 3 ð14cÞ alternative is to use cR0 ¼0.045, c0S1 ¼ DcS1 ¼0.042 0.025¼0.017,
where DcSi  c0Si cSi
(i.e., DcSi signifies the additional cost of CCS c0S2 ¼ DcS2 ¼0.034 0.022¼0.012, c0S3 ¼ DcS3 ¼ 0.032 0.020¼0.012,
0
source Si with reference to its non-CCS option Si) with K1, K2, and cS1 ¼0.0, cS2 ¼0.0 and cS3 ¼0.0. These modified cost factors allow
K3 denoting constants. Eq. (14a) is obtained because of the easier validation during the 28 LP runs (for positive and negative
constancy of the first term (noting FR0 ¼400,000 TJ), the third deviations of the 14 rows in Table 6) for the optimal cases listed in
term (which is the cost of total available sources) and the fourth the last two columns of Table 6 (although the actual objective
term (which is zero for no excess energy) in Eq. (3b). Eq. (14a) function value will differ from that obtained on using base cost
gives Eq. (14b) on expanding the summation and absorbing the factors). Note that two new sub-cases (in addition to the eight
DcS3F 0S3 term into the constant, whereas Eq. (14b) finally yields primary cases in Table 5) are discovered during the LP runs, and
Eq. (14c) on substituting for F 0S2 from Eqs. (13a) and (13b). For the tabulated as Case B2S and Case C1S at the bottom of Table 6. These
introduction of S10 (F 0S1 40) to be advantageous, the operating sub-cases are variations of Case B2 (Fig. 5b) and Case C1 (Fig. 6a) due
cost must decrease. This requires the term in square brackets in to the pinch point jumping from 105 to 75 t CO2/TJ. Such pinch
Eq. (14c) be negative, giving the following simple condition: jumps have been observed earlier and are discussed by Shenoy and
Bandyopadhyay (2007).
DcS1 o RC DcS2 where RC ¼ ð55210Þ=ð75215Þ ð14dÞ
The cost criterion in Eq. (15a) may be rewritten as
Eq. (14d) specifies the condition where it is optimal to
ðc0Si 2cSi Þ=ðC Ei 2C 0Si Þ o ðc0Sði þ 1Þ 2cSði þ 1Þ Þ=ðC Eði þ 1Þ 2C 0Sði þ 1Þ Þ ð15bÞ
introduce S10 to partially replace S20 . In other words, the optimal
solution is Case C3 for a positive deviation (i.e., DcS1 ¼ RCDcS2 þ d) In this slightly different version of the cost criterion, the
and it is Case D for a negative deviation (i.e., DcS1 ¼RCDcS2  d), quantity (c0Si  cSi)/(CEi  C 0Si ) may be called the prioritized cost
where d is a small deviation (say, 10  6) of DcS1 from RCDcS2. The (Fraser et al., 2005; Shenoy and Bandyopadhyay, 2007) of the
above result is given in the last row of Table 6, where several ith CCS source. Note that it is not necessary to maximize the usage
other results are tabulated based on a generalization of Eq. (14d) of the CCS source with the minimum unit cost (i.e., cost per
as given next in Eq. (15a). energy unit) in order to minimize the overall operating cost. The
For the optimal introduction of CCS source Si0 as a partial prioritized cost, in a sense, is a ratio of the economic/operating
replacement for CCS source S(iþ1)0 , the general cost criterion may cost change to the environmental benefit change when the CCS

Table 6
Optimality analysis for minimum operating cost based on the RC cost criterion.

Replace RC DcS1 DcS2 DcS3 Forbidden Optimal for Optimal for


þ ve deviation  ve deviation

R0 by S10 (55  10)/(75  0) RC DcR0 7 d S20 , S30 Case A Case B1


R0 by S20 pinch at 105 (75  15)/(105  0) RC DcR0 7 d S10 , S30 Case B2S Case B2
R0 by S20 pinch at 75 (75  15)/(75  0) RC DcR0 7 d S10 , S30 Case A Case B2S
R0 by S30 (75  20)/(75  0) RC DcR0 7 d S10 , S20 Case A Case B3
R0 by S20 pinch at 105 (75  15)/(105  0) RC DcR0 7 d S30 Case C1S Case C1
R0 by S20 pinch at 75 (75  15)/(75  0) RC DcR0 7 d S30 Case B1 Case C1S
R0 by S30 (75  20)/(75  0) RC DcR0 7 d S20 Case B1 Case C2
S20 by S10 (55  10)/(75  15) RC DcS2 7 d S30 Case B2 Case C1
S20 by S30 (105  20)/(75  15) RC DcS2 7 d S10 Case B2 Case C3
S30 by S10 (55  10)/(75  20) RC DcS3 7 d S20 Case B3 Case C2
S30 by S20 (75  15)/(75  20) RC DcS3 7 d S10 Case B3 Case C3
S20 by S30 (105  20)/(75  15) 0.009  d RC DcS2 7 d Case C1 Case D
S30 by S20 (75  15)/(75  20) 0.009  d RC DcS3 7 d Case C2 Case D
S20 by S10 (55  10)/(75  15) RC DcS2 7 d Case C3 Case D

Note: DcR0 ¼0.045, DcS1 ¼ 0.017, DcS2 ¼0.012, DcS3 ¼ 0.012; d ¼ 10  6.


Case B2S: FR0 ¼500,000 TJ, F 0S1 ¼0, F 0S2 ¼391,667 TJ, F 0S3 ¼ 0.
Case C1S: FR0 ¼500,000 TJ, F 0S1 ¼200,000 TJ, F 0S2 ¼ 241,667 TJ, F 0S3 ¼0.
A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327 325

option is used over the non-CCS one. The prioritized costs provide Table 7
an order in which the CCS sources may be used; however, they MILP validation by varying cost data.
may be difficult to calculate a priori in some multi-pinch and
Input cost data
pinch-jump cases.
cR0 0.045 0.045 0.045 0.045 0.045 0.045
c0S1 0.042 0.042 0.036 0.036 0.042 0.036
7. MILP formulation and validation for total cost optimization c0S2 0.0337 0.034 0.034 0.034 0.034 0.034
c0S3 0.032 0.032 0.033 0.032 0.032 0.032
cS1 0.025 0.025 0.032 0.0315 0.025 0.0315
The LP formulation discussed so far focuses on the optimiza- cS2 0.022 0.022 0.022 0.022 0.0164 0.0196
tion of only the operating cost. In this section, the formulation is cS3 0.020 0.020 0.020 0.020 0.0141 0.016
extended to a mixed integer linear program (MILP) for the c0F S1 1000 1000 1000 1000 1000 1000
optimization of the total cost given by the following expression: c0F S2 800 800 800 800 800 800
c0F S3 700 700 700 700 700 400
NR
X NS
X NS
X NS X
X ND
TC ¼ cRk F Rk þ c0Si F 0Si þ cSi ðF Si f iE Þ þ c0F Si y0ij ð16aÞ Optimum solution from MILP
k¼0 i¼1 i¼1 i¼1j¼1 FR0 (104 TJ) 40 40 40 40 40 40
F 0S1 (104 TJ) – – 20 20 – 20
In Eq. (16a), the first three terms are identical to those in F 0S2 (104 TJ) 56.6667 – 41.6667 – 42.5 27.5
Eq. (3a). The last (additional fourth) term corresponds to the fixed F 0S3 (104 TJ) – 56.3636 – 40 10 10
cost with y0ij denoting the binary variables (1 if demand j is met by Case B2 B3 C1 C2 C3 D
CCS source i, and 0 otherwise) and c0F Si denoting the associated
fixed cost factors. The binary variable associated with each CCS
Based on the preceding discussion, Table 7 summarizes the
match requires the following big-M disjunctive constraint:
MILP validation for six cases (corresponding to zero excess, i.e.,
0
f ij My0ij r0 ð16bÞ FR0 ¼400,000 TJ) in Table 5 by varying the cost data. Case B3 is the
optimum for the base-case cost data in Table 1 (when c0S2 ¼0.034),
where M is a very large constant value (providing the upper limit whereas Case B2 is the optimum when c0S2 ¼ 0.0337 (i.e.,
of fij0 ). Depending on whether the binary variable is zero or unity, c0S2 o0.03376 in accordance with Eq. (17d)). The last four columns
Eq. (16b) ensures that the associated continuous variable is either in Table 7 show possibilities where Cases C1, C2, C3 and D are
zero or positive. optimum when cost data are varied (with changed values high-
The objective is to minimize the total cost TC in Eq. (16a) subject lighted in bold).
to the constraints given by Eqs. (4), (5), (8), and (16b). The MILP
formulation can be solved using optimization software such as GAMS.
On using the cost data in Table 1 along with the earlier-specified fixed 8. Conclusions
cost factors of 1000, 800, and 700 (originally from Pekala et al., 2010),
the optimum for minimum total cost corresponds to the values of A detailed mathematical formulation has been developed for
FR0 ¼400,000 TJ and F 0S3 ¼563,636 TJ. This validates the result the targeting of the energy allocation problem with CCS. For the
obtained earlier in Table 5 and reported as Case B3. case of CCS with zero excess, it has been shown that CCS sources
Furthermore, the MILP may be used to validate other cases in can be treated as resources with their cost per energy unit taken
Table 5 by simply varying the cost data as demonstrated below. as the increment over the non-CCS option. The limiting composite
Based on Eq. (16a), the expression for the operating cost in Eq. (14a) is a valuable, unique curve for targeting with and without CCS. It
may be extended to obtain the following equation for the total cost: is readily obtained using a single computation of the composite
NS
X NS X
X ND table algorithm (CTA) by considering demands and only non-CCS
TC ¼ DcSi F 0Si þK 1 þ c0F Si y0ij ð17aÞ sources with their total energy availabilities.
i¼1 i¼1j¼1
In this work, the recent methodology (Shenoy, 2010) for
Applying Eq. (17a) for Cases B2 and B3 (using values for F 0Si targeting and design of carbon emission networks has been
from Table 5 along with fixed cost factors of 800 for CCS—Oil extended by the inclusion of carbon capture and storage (CCS)
match and 700 for CCS—Coal match) gives of the fossil fuel sources. As in the earlier work (Shenoy, 2010),
the CTA is applied to generate the limiting composite and the
TC B2 ¼ 566667 DcS2 þ K 1 þ 800 for Case B2 ð17bÞ NNA is used to design the network. However, the earlier work was
limited to targeting of clean-carbon resources with no CCS
TC B3 ¼ 563636 DcS3 þ K 1 þ 700 for Case B3 ð17cÞ
sources, whereas the present work focuses on targeting of
The optimum is Case B3 for the cost data in Table 1 because clean-carbon resources with multiple (specifically, one, two or
TCB3 yields the minimum value. For Case B2 to be the optimum, it three) CCS sources. Further, the earlier work did not consider any
is necessary (though not sufficient) that TCB2 oTCB3. On substitut- cost data nor discuss any cost analysis. This work develops a
ing from Eqs. (17b) and (17c) formal mathematical formulation, discusses detailed cost optimi-
zation, and considers the effects of varying cost data. Broadly, the
566667DcS2 þ800 o 563636DcS3 þ 700 ð17dÞ
contribution takes a systems engineering approach to energy
Note that Eq. (17d) is an extension of the Rc cost criterion utilization with regeneration through carbon sequestration for
(Eqs. (14d) and (15) considering only operating cost) to the case of sustainable development. The contribution is inspired by the key
total cost. Eq. (17d) yields two possibilities for Case B2 to be representation in Fig. 2 of a CCS source and its corresponding non-
optimum: DcS2 o(563,636 DcS3  100)/566,667 (i.e., DcS2 o0.01176 CCS source in terms of an equivalent composite. The equivalent
or C 0S2 o0.03376 on substituting DcS3 ¼0.012 and cS2 ¼0.022); and composite representation has two important implications: only
DcS3 4(566,667 DcS2 þ100)/563,636 (i.e., DcS3 40.01224 or non-CCS sources (with energy values given in terms of their
c0S3 40.03224 on substituting DcS2 ¼0.012 and cS3 ¼0.020). Both specified totals) are first considered to establish a unique limiting
possibilities are readily verified by solving an MILP to yield Case B2 composite; and only CCS sources (depicted as line segments
as the optimum (FR0 ¼400,000 TJ and F0 S2 ¼566,667 TJ) on choosing between CSi0 and CSi) are then regarded as resources for targeting
different values for the cost factors (c0S2 o0.03376 or c0S3 40.03224). by profile matching. The approach is novel because it allows
326 A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327

targets to be set and networks to be synthesized even when cost structure. A significant advantage of the CTA/NNA methodology
data are unavailable or uncertain. The results have been validated is that targets may be set and networks synthesized based on
through an LP formulation for minimum operating cost and an primary cases even in the absence of cost data.
MILP formulation for minimum total cost considering the sensi-
tivity of varying cost data.
The presented approach differs from earlier established methods Nomenclature
in the following ways. Firstly, the presented approach for carbon
emission networks requires handling of multiple CCS sources and C emission factor or carbon intensity, t CO2/TJ
consequently determination of Ns unknown variables corresponding c operating cost (factor) per energy unit, cost unit/TJ
to the optimum requirements of Ns CCS sources. In contrast, cF fixed cost factor
established methods for water and hydrogen networks typically CCS carbon capture and storage/sequestration
handle a single regeneration stream and consequently are a special CEN carbon emission network
case of the presented approach for Ns ¼1. Secondly, in carbon CTA composite table algorithm
emission networks, the operating cost in the objective function D demand for energy in a region/sector
depends on the clean resources, the sources with CCS, and the E excess (or unused) energy, TJ
sources without CCS (excluding the excess or unused energy). In F energy, TJ
water and hydrogen networks, the operating cost depends on the f energy allocation for cell matches in matching matrix, TJ
resources (freshwater usage and hydrogen makeup, respectively) i index denoting number of source (CCS or non-CCS)
and the regeneration, but not on the sources (which are internal). j index denoting number of demand
Thirdly, in carbon emission networks, there is a constraint on the K1, K2, K3 constants in Eq. (14)
CCS and non-CCS portions of each source totaling the specified k index denoting number of clean-carbon resource
availability. In water and hydrogen networks, no such constraint LP linear programming
exists except that the regeneration flowrate must not exceed the MILP mixed integer linear programming
clean resource flowrate. Finally, in carbon emission networks, the M emission load or carbon footprint, t CO2 or constant in
‘equivalent concentration’ for inlet regeneration is known because it big-M constraint
is given by the specified non-CCS source emission factor Csi. In water NNA nearest neighbors algorithm
and hydrogen networks, the inlet regeneration concentration is not N total number
known and must be determined. OC operating cost
In this work, multiple CCS sources along with clean-carbon R resource (zero-carbon or low-carbon)
resources are simultaneously targeted by profile matching with R0 resource (zero-carbon)
the limiting composite. It may be noted that Figs. 3, 5, 6, and 7 RC ratio of emission-factor differences defined in eq 15a
show ‘utility pinches’ where the target profile touches the limit- S source for energy (conventional fossil fuel such as coal,
ing composite. The problem of targeting CCS sources is a challen- oil, or natural gas)
ging one and its solution by profile matching is non-trivial. For UTA unified targeting algorithm
this purpose, fundamental carbon emission load balance equa- y binary variable associated with total cost term
tions may be utilized where clean-carbon resources are repre- D1 net system energy deficit, TJ
sented by arrows and CCS sources by line segments. Targeting D2 net system emission load deficit, t CO2
typically results in 2NS optimal primary cases for NS CCS sources Dc additional cost of CCS source with reference to its non-
and does not require cost data. CCS option
Design of carbon emission networks (CENs) that achieve the
already-established targets may be next done by systematically Subscripts
using the nearest neighbors algorithm (NNA) along with the
matching matrix representation. It must be emphasized that the D demand
NNA, by merely varying the order of satisfaction of demands, E excess (or unused) energy
allows synthesis of many optimum CENs (all meeting the targets i index for source (CCS or non-CCS)
and the specified emission load limits) for each primary case. j index for demand
Cost analysis may be now performed in terms of both operat- k index for clean-carbon resource
ing cost as well as total cost, if cost data are available. The cost max maximum (specified availability of clean-carbon
ranges in which the different primary cases are optimal may be resource)
determined using the RC cost criterion. R clean-carbon resource (zero-carbon or low-carbon)
Optimization may be finally done by solving a linear program- R0 resource (zero-carbon)
ming (LP) formulation for minimum operating cost. The solutions P pinch
by mathematical programming permit validation of the earlier S source (CCS or non-CCS)
targeting results and network designs. They also help in identi- total total (availability of source)
fication of unusual new sub-cases (of pinch jumps, for instance)
and additional networks. Superscript
To summarize, the methodology in this work comprises four
distinct stages: targeting using the CTA/UTA along with funda- 0
prime to denote CCS source
mental balances for profile matching with the limiting composite
curve; designing networks for each of the optimal primary cases
using the NNA; analyzing costs using the RC cost criterion; and
finally optimizing solutions using mathematical programming References
formulations. This work therefore proposes a hybrid approach
based on the CTA/NNA as well as mathematical programming Agrawal, V., Shenoy, U.V., 2006. Unified conceptual approach to targeting and
design of water and hydrogen networks. AIChE J. 52 (3), 1071–1082.
(LP/MILP). The methodology using the CTA/NNA provides deeper Atkins, M.J., Morrison, A.S., Walmsley, M.R.W., 2008. Carbon emissions pinch
insights into the problem, its pinch points and its network analysis (CEPA) for emissions reduction in the New Zealand electricity sector.
A.U. Shenoy, U.V. Shenoy / Chemical Engineering Science 68 (2012) 313–327 327

In: Proceedings of the Society of Chemical Engineers New Zealand Annual Lee, S.C., Ng, D.K.S., Foo, D.C.Y., Tan, R.R., 2009. Extended pinch targeting
Conference (SCENZ08), New Zealand. techniques for carbon-constrained energy sector planning. Appl. Energy 86,
Atkins, M.J., Morrison, A.S., Walmsley, M.R.W., 2009a. Carbon emissions pinch 60–67.
analysis (CEPA) for emissions reduction in the New Zealand electricity sector. Pekala, L.M., Tan, R.R., Foo, D.C.Y., Jezowski, J.M., 2010. Optimal energy planning
Appl. Energy 87, 982–987. models with carbon footprint constraints. Appl. Energy 87 (6), 1903–1910.
Atkins, M.J., Morrison, A.S., Walmsley, M.R.W., 2009b. Carbon emissions pinch Prakash, R., Shenoy, U.V., 2005a. Targeting and design of water networks for fixed
analysis (CEPA) for emissions reduction in the New Zealand electricity sector. flowrate and fixed contaminant load operations. Chem. Eng. Sci. 60 (1),
Chem. Eng. Trans. 18, 261–266.
255–268.
Bandyopadhyay, S., Malik, R.K., Shenoy, U.V., 1999. Invariant rectifying-stripping
Prakash, R., Shenoy, U.V., 2005b. Design and evolution of water networks by
curves for targeting minimum energy and feed location in distillation. Comput.
Chem. Eng. 23, 1109–1124. source shifts. Chem. Eng. Sci. 60, 2089–2093.
Bandyopadhyay, S., Mishra, M., Shenoy, U.V., 2004. Energy-based targets for Shenoy, U.V., 1995. Heat Exchanger Network Synthesis: Process Optimization by
multiple-feed distillation columns. AIChE J. 50, 1837–1853. Energy and Resource Analysis Gulf Publishing, Houston, TX.
Crilly, D., Zhelev, T., 2008. Emissions targeting and planning—an application of Shenoy, U.V., 2010. Targeting and design of energy allocation networks for carbon
CO2 emissions pinch analysis (CEPA) to the Irish electricity generation sector. emission reduction. Chem. Eng. Sci. 65 (23), 6155–6168.
Energy 33, 1498–1507. Shenoy, U.V., 2011. Unified targeting algorithm for diverse process integration
Crilly, D., Zhelev, T., 2009. Expanded emissions and energy targeting: a further problems of resource conservation networks. Chem. Eng. Res. Des..
application of CO2 emissions pinch analysis (CEPA) to the Irish electricity doi:10.1016/j.cherd.2011.04.021.
generation sector. Chem. Eng. Trans. 18, 75–80. Shenoy, U.V., Bandyopadhyay, S., 2007. Targeting for multiple resources. Ind. Eng.
Das, A.K., Shenoy, U.V., Bandyopadhyay, S., 2009. Evolution of resource allocation Chem. Res. 46 (11), 3698–3708.
networks. Ind. Eng. Chem. Res. 48 (15), 7152–7167. Shenoy, U.V., Sinha, A., Bandyopadhyay, S., 1998. Multiple utilities targeting for
Deng, C., Feng, X., 2009. Optimal water network with zero wastewater discharge in heat exchanger networks. Chem. Eng. Res. Des. 76, 259–272.
an alumina plant. WSEAS Trans. Environ. Dev. 2 (5), 146–156. Shethna, H.K., Singh, H., Makwana, Y., Castillo, F.J.L., Shenoy, U.V., 1999. Multiple
El-Halwagi, M.M., 1997. Pollution Prevention through Process Integration: Sys- utilities optimization to improve process economics. Petrol. Technol. Q.
tematic Design Tools Academic Press, San Diego, CA. Autumn, 133–139.
El-Halwagi, M.M., 2006. Process Integration Elsevier, Amsterdam. Smith, R., 2005. Chemical Process: Design and Integration Wiley, NY.
El-Halwagi, M.M., Manousiouthakis, V., 1990. Simultaneous synthesis of mass- Steeneveldt, R., Berger, B., Torp, T.A., 2006. CO2 capture and storage: closing the
exchange and regeneration networks. AIChE J. 36, 1209–1219.
knowing-doing gap. Chem. Eng. Res. Des. 84, 739–763.
Foo, D.C.Y., Tan, R.R., Ng, D.K.S., 2008. Carbon and footprint-constrained energy
Tan, R.R., Foo, D.C.Y., 2007. Pinch analysis approach to carbon-constrained energy
planning using cascade analysis technique. Energy 33, 1480–1488.
Fraser, D.M., Howe, M., Hugo, A., Shenoy, U.V., 2005. Determination of mass sector planning. Energy 32, 1422–1429.
Wall, T.F., 2007. Combustion process for carbon capture. Proc. Combust. Inst. 31,
separating agent flows using the mass exchange grand composite curve.
Chem. Eng. Res. Des. 83, 1381–1390. 31–47.
Hallale, N., Fraser, D.M., 1998. Capital cost targets for mass exchange networks. Wang, Y.P., Smith, R., 1994. Wastewater minimization. Chem. Eng. Sci. 49,
A special case: water minimisation. Chem. Eng. Sci. 53 (2), 293–313. 981–1006.
Kemp, I.C., 2007. Pinch Analysis and Process Integration: A Users Guide on Yang, H., Xu, Z., Fan, M., Gupta, R., Slimane, R.B., Bland, A.E., Wright, I., 2008.
Process Integration for the Efficient Use of Energy Butterworth-Heinemann, Progress in carbon dioxide separation and capture: a review. J. Environ. Sci. 20,
Oxford, UK. 14–27.

You might also like