Designing optimal bioethanol networks with purification for integrated biorefineries

Energy Conversion and Management xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Energy Conversion and Management journal homepage: www.elsevier.com/locate/enconman

Designing optimal bioethanol networks with purification for integrated biorefineries Akshay U. Shenoy a, Uday V. Shenoy b,⇑ a b

Indian Institute of Management, IIMB, Bannerghatta Road, Bilekahalli, Bangalore 560 076, India Synew Technologies, A 502, Galleria, Hiranandani Gardens, Powai, Mumbai 400 076, India

a r t i c l e

i n f o

Article history: Available online xxxx Keywords: Bioethanol Biorefinery Biomass utilization Sustainable design Systems engineering Optimization Process integration Resource conservation Pinch analysis

a b s t r a c t Bioethanol networks with purification for processing pathways in integrated biorefineries are targeted and designed in this work by an analytical approach not requiring graphical constructions. The approach is based on six fundamental equations involving eight variables: two balance equations for the stream flowrate and the bioethanol load over the total network system; one equation for the above-pinch bioethanol load being picked up by the minimum fresh resource and the purified stream; and three equations for the purification unit. A solution strategy is devised by specifying the two variables associated with the purifier inlet stream. Importantly, continuous targeting is then possible over the entire purifier inlet flowrate range on deriving elegant formulae for the remaining six variables. The Unified Targeting Algorithm (UTA) is utilized to establish the minimum fresh bioethanol resource flowrate and identify the pinch purity. The fresh bioethanol resource flowrate target is shown to decrease linearly with purifier inlet flowrate provided the pinch is held by the same point. The Nearest Neighbors Algorithm (NNA) is used to methodically synthesize optimal networks matching bioethanol demands and sources. A case study of a biorefinery producing bioethanol from wheat with arabinoxylan (AX) coproduction is presented. It illustrates the versatility of the approach in generating superior practical designs with up to nearly 94% savings for integrated bioethanol networks, both with and without process constraints, for grassroots as well as retrofit cases. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Process integration tools broadly aim at minimizing external resource requirements through maximizing internal material reuse/ recycle and energy recovery [1–3]. In turn, they endeavor to minimize adverse environmental impact and maximize profitability through sustainable process designs [4,5]. Process integration using pinch analysis as well as mathematical programming has been applied to various biorefinery configurations [6–10] including value added production pathways and combined heat and power generation [11,12]. However, as pointed out by Martinez-Hernandez et al. [13] in their seminal work, biorefinery mass integration at the product level for potential utilization of various products within the biorefinery processes to reduce material utilities and/or feedstocks has not been explored. Thus, targeting as in pinch analysis for the minimum utility [14–16] is valuable for screening and scoping of different product

⇑ Corresponding author. Tel.: +91 22 4010 4615. E-mail address: [email protected] (U.V. Shenoy).

allocation networks as more complex and advanced process technologies emerge in biorefining. Biofuels, being renewable or green fuels, potentially provide a sustainable way to satisfy the world’s ever-increasing energy demands [17,18]; however, their production processes must be cost-effective and process designs optimal in terms of efficient use of resources. Bioethanol, a gasoline additive/substitute, is by far the most widely used biofuel for transportation worldwide [19,20]. Recently, Martinez-Hernandez et al. [13] have emphasized the need to develop new tools for the integrated processing of starch and lignocellulosic feedstocks in bioethanol production, wherein ethanol can be used as utility for biomass fractionation or pretreatment as well as chemical reactant. Their methodology, adapted from hydrogen pinch analysis [21], seeks to minimize the bioethanol requirement within the biorefinery using a graphical approach based on composite curves and a surplus diagram. However, the surplus diagram method requires transferring of data from one plot to another and is iterative [22]. Their analysis for bioethanol network design thus involves graphical construction, tedious calculation and typically several iterations; so, it is adapted to a spreadsheet tool using Excel–VBA. In general, graphical

http://dx.doi.org/10.1016/j.enconman.2014.02.066 0196-8904/Ó 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

2

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

targeting methodologies can often be unwieldy although graphical representations in process integration can provide valuable visual insights. A purely analytical method offers various advantages including easy implementation to real-world problems with big data sets, speedy what-if analysis and high accuracy. The aim of this work is to develop a straightforward non-iterative methodology to continuously target and design bioethanol networks over the entire purifier inlet flowrate range. The totally analytical approach proposed here requires no graphical constructions and is based on two versatile algorithms along with a set of six fundamental equations. This work thus extends the unified conceptual approach developed by Agrawal and Shenoy [23] for water and hydrogen management to biorefinery integration [24]. It involves a systems engineering methodology for the holistic understanding of the flow of a species within a process to determine its optimal allocation between sources and demands. The overall goal is to optimally allocate resources (e.g., water, hydrogen, energy, and component species) that have both a quantity (load) and a quality (level) by matching demands and sources (after appropriate mixing, if necessary). In the present context, the flow of bioethanol in a biorefinery is studied to target the minimum makeup fresh resource, and optimal bioethanol networks are designed with and without purification under known process constraints. Here, the classic two-stage approach of pinch analysis [25,26] is followed: first, minimum bioethanol flow targets are established by the Unified Targeting Algorithm (UTA); and, second, optimal bioethanol networks are systematically designed by the Nearest Neighbors Algorithm (NNA). The major advantage of the UTA [27] and the NNA [28] is that they both provide a unified methodology. The UTA is applicable to a diverse range of process integration problems [29], including those of heat/mass exchange, water, hydrogen, nitrogen, oxygen, carbon emission, and property-based material reuse networks. Similarly, the NNA has been extensively applied for the synthesis of various resource conservation networks [2] including water networks [30,31], hydrogen networks [23], carbon emission networks [32,33], and material reuse networks [34]. Some practical applications of the NNA include industrial water conservation in a steel plant [35], zero wastewater discharge in an alumina plant [36], ultrapure water recovery scheme for wafer fabrication section in a semi-conductor plant [37], sustainable energy planning using agricultural-land/water/ carbon footprint [38], and property integration in a metal degreasing process as well as hydrogen integration in a petroleum refinery [39]. In what follows, fundamental equations for bioethanol networks with purification are presented and used to study the continuous variation of the optimal targets as a function of the purifier inlet flowrate. The UTA is utilized to determine the minimum fresh bioethanol flowrate target and the pinch. The NNA is then fruitfully used for designing networks to meet the targets with and without a purifier as well as with and without process constraints. A case study is analyzed of a complex biorefinery with arabinoxylan (AX) extraction, wherein ethanol is a biorefinery product as well as a process stream, resulting in demands and sources at different purity levels. Finally, a graphical explanation of the targeting methodology is provided. 2. Methodology

F R  F E ¼ D1

where D1 

F R yR  F E yE ¼ D2

X

Fd 

where D2 

X

X

Fs

F d yd 

ð1aÞ X

F s ys

ð1bÞ

where F denotes stream flowrate, y denotes bioethanol component purity fraction, and subscripts R and E denote fresh resource and excess/waste, respectively. The net system flowrate (D1) and the net system component load (D2) are obtained as deficits by taking the sum of all demands (denoted by subscript d) and subtracting the sum of all sources (denoted by subscript s). Defining such net quantities (D) that are constant for a given network system is advantageous in resource optimization [40]. For the purification unit, the inlet stream entering at purity yin is considered as a demand and the streams leaving at ypr (product stream of high purity) and yr (residue stream of low purity that typically goes to waste) as two sources [41]. For a purification process starting at an inlet purity yin and achieving an outlet purity ypr (at or above the pinch purity yp), the component load balance over the above-pinch region [23] yields

F R ðyR  yp Þ þ F pr ðypr  yp Þ ¼ M p

ð2Þ

where M denotes component load, and subscripts p and pr denote pinch and product stream of high purity, respectively. For the purifier, the flowrate and bioethanol component load balances are simply given by

F in ¼ F pr þ F r F in yin ¼ F pr ypr þ F r yr

ð3aÞ ð3bÞ

Further, the product purity ypr and the component recovery R, as defined below, are usually specified for the purifier:

R ¼ F pr ypr =ðF in yin Þ

ð3cÞ

Since Eqs. (1)-(3) constitute six equations in eight unknowns (FR, FE, Fin, Fpr, Fr, yE, yin and yr), there are two degrees of freedom. Therefore, two variables may be specified and the equations then solved to determine the targets as discussed next. 2.2. Solution strategy for continuous targeting Let the two variables corresponding to the purifier inlet (i.e., Fin and yin) be specified. Then, the equations may be written in terms of Fin, yin and other known quantities (D1, D2, yR, yp, Mp, ypr and R) as follows. Eq. (2) is rearranged using Eq. (3c) to establish the minimum flowrate target for the fresh bioethanol resource as

F R ¼ F R0  KF in

ð4aÞ

where FR0 = Mp/(yR  yp) and K = (R yin/ypr)(ypr  yp)/(yR  yp). Notably, the form of Eq. (4a) specifies FR0 as the fresh bioethanol flowrate target in the absence of a purifier. Eqs. (1) and (4a) then yield the flowrate and purity of the excess/waste as

F E ¼ F R0  D1  KF in

ð4bÞ

yE ¼ ðF R0 yR  D2  KyR F in Þ=ðF R0  D1  KF in Þ

ð4cÞ

Eqs. (3a)–(3c) may be combined to obtain the outlet flowrates and the residue purity for the purifier as

F pr ¼ ðRyin =ypr ÞF in

ð4dÞ

F r ¼ ð1  Ryin =ypr ÞF in

ð4eÞ

2.1. Fundamental equations for bioethanol networks with purification

yr ¼ yin ð1  RÞ=ð1  Ryin =ypr Þ

ð4fÞ

The fundamental balance equations for stream flowrate and component load [23] over the total network system with purification (Fig. 1) are

As per Eq. (4a), the fresh bioethanol resource flowrate target FR varies (decreases) linearly with the purifier inlet flowrate Fin provided the pinch is held by the same point [i.e., (Mp, yp) and

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

3

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

Fig. 1. Schematic representation of bioethanol allocation network with purification.

consequently (FR0, K) are constant]. Further, Eq. (4f) suggests that the residue purity yr does not depend on the purifier inlet flowrate Fin but depends only on the purifier inlet purity yin. Importantly, Eqs. (4) provide elegant formulae for establishing continuous targets over the Fin range as illustrated through the case study later. Furthermore, Eqs. (4a)–(4c) provide apposite formulae for the targets in the case of no purification on simply setting Fin to zero. The determination of the fresh bioethanol flowrate target FR from Eq. (4a) requires knowledge of the pinch point. The next subsection presents the UTA to analytically determine the pinch and establish optimal flowrate targets. 2.3. Unified Targeting Algorithm (UTA)

2. Tabulate in the second column the algebraic sum of the corresponding flows of demands and sources present at each level l using

Fl ¼

X

1. Arrange in the first column all levels of purity y (without repetition) for demands and sources in decreasing order. The levels define the limits of the various intervals (e.g., the first interval goes from the first level to the second level). Include an arbitrarily chosen purity (say, 0%) with zero flow so that an extra level entry appears at the bottom of the first column.

X

F s;l

ð5aÞ

Demand flows are taken as positive and source flows as negative, adopting the positive demands sign convention without any loss of generality. 3. Tabulate in the third column the cumulative flows based on the previous (second) column. This yields the net flow deficit/surplus for each interval by summing the flows up to the previous level in accordance with

F net;l ¼

In the UTA [29], all inlet streams are treated as demands (to processes) and all outlet streams as sources (to other processes). Although the graphical representation related to the UTA for allocation network problems is the limiting composite curve (LCC, as discussed later), no plots are actually required for targeting as highlighted in the analytical method below. The main steps of the UTA are concisely given below and may be implemented through a compact table (as in Tables 2 and 4). Additional details, including the rationale behind the steps, are provided by Shenoy [27].

F d;l 

X

F l ¼ F net;l1 þ F l1

ð5bÞ

where Fnet,1 = 0 and the summation goes from 1 to (l  1). The net flow in an interval corresponds to the reciprocal of the slope of the associated segment on the LCC. 4. Tabulate in the fourth column the net load for each interval by multiplying the net flow (from the previous step) and the level difference (from the first column) for that interval as per the following formula:

Mnet;l ¼ F net;l ðyl1  yl Þ

ð5cÞ

where Mnet,1 = 0. These values signify the net bioethanol load deficit/surplus in each interval, after complete allocation within the interval. 5. Tabulate in the fifth column the cumulative load by summing the net loads (from the previous column) up to that level according to

Ml ¼

X

Mnet;l ¼ M l1 þ Mnet;l

ð5dÞ

Table 2 Implementation of UTA without process constraints. Purity level y

Flow at level F (t/y)

Net flow Fnet (t/y)

Net load Mnet (t/y)

Cumul. load M (t/y)

Fresh flow at Fin = 0 FR (t/y)

Fresh flow at Fin = 26,709 FR (t/y)

0.96 0.9152 0.7 0.6822 0.64 0.5095 0.1509 0.0249 0

22,567 21,670 57,075 55,633 892 2049 1437 80,304 0

0 22,567 897 57,972 2339 1447 602 2039 82,343

0 1011.0016 193.0344 1031.9016 98.7058 188.8335 215.8772 256.9140 2050.3407

0 1011.0016 1204.0360 2235.9376 2334.6434 2523.4769 2307.5997 2050.6857 0.345

0 12512.396 4067.689 7125.359 6557.987 5187.003 2730.564 2111.714 0.346

0 2199.186 12270.685 9341.338 10161.663 12037.201 15077.688 15799.346 17927.953

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

4

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

Table 4 Implementation of UTA with process constraints. Purity level y

Flow at level F (t/y)

Net flow Fnet (t/y)

Net load Mnet (t/y)

Cumul. load M (t/y)

Fresh flow at Fin = 0 FR (t/y)

Fresh flow at Fin = 55,633 FR (t/y)

0.96 0.9152 0.7 0.6822 0.64 0.5095 0.1509 0.0249 0

22,567 21,670 57,075 0 892 2049 1437 80,304 0

0 22,567 897 57,972 57,972 57,080 55,031 53,594 26,710

0 1011.0016 193.0344 1031.9016 2446.4184 7448.9400 19734.1166 6752.8440 665.0790

0 1011.0016 1204.0360 2235.9376 4682.3560 12131.2960 31865.4126 38618.2566 37953.1776

0 12512.396 4067.689 7125.359 13152.685 24935.860 37706.085 39767.538 38105.600

0 8969.158 29963.778 27173.395 21672.948 10940.716 612.985 2460.297 762.451

where M1 = 0 and the summation goes from 1 to l. The LCC may be obtained by simply plotting the y (first column) vs. M (fifth column) data. 6. Tabulate in the next two columns (for two limiting values of Fin) the fresh bioethanol flowrates for different (Ml, yl) value-pairs using the following formula based on Eq. (4a):

F R;l ¼ ½M l  ðRyin =ypr Þðypr  yl ÞF in =ðyR  yl Þ

ð5eÞ

where Ml is the load (fifth column) and yl is the purity level (first column). The highest value from Eq. (5e) specifies the minimum fresh bioethanol resource flowrate target FR with the purity level yp (in the first column) and the load Mp (in the fifth column) corresponding to it defining the pinch. Eqs. (4) may be then utilized to determine all the targets (for FR, FE, yE, Fpr, Fr, and yr). The first five columns form the basic UTA table, whereas the last two columns provide the extension for the continuous targets in integrated biorefineries. The Nearest Neighbors Algorithm (NNA) is described next to design bioethanol networks that meet the targets established.

2.4. Nearest Neighbors Algorithm (NNA) Consider a general allocation network problem with n sources (S1 to Sn) and m demands (D1 to Dm). The fresh bioethanol resource and two purifier outlet streams are included as sources, whereas the purifier inlet stream and excess/waste are taken as demands. Biorefineries involve fixed flowrate processes; therefore, the bioethanol networks are synthesized by the original NNA [28] and not the Enhanced NNA [42], which is applicable to fixed load processes. To synthesize a biorefinery allocation network that satisfies the target at a particular value of Fin, the basic steps for the NNA are outlined below in brief. Further details, including explanations of equations used, are given by Shenoy and Shenoy [43].

3. Nearest neighbor sources To fulfill a demand Dp according to the principle of nearest neighbors [28], choose two sources Sh and Sl such that Sh has a level just higher than that of Dp, and Sl has a level just lower than that of Dp. To calculate their required amounts, go to Step 4. 4. Two-source calculation Calculate the amounts of the higher-level source Sh and the lower-level source Sl necessary to meet the demand Dp from

F Sh;Dp ¼ F Dp ðyDp  ySl Þ=ðySh  ySl Þ

ð6aÞ

F Sl;Dp ¼ F Dp ðySh  yDp Þ=ðySh  ySl Þ

ð6bÞ

If the amounts FSh,Dp and FSl,Dp calculated from Eqs. (6) are less than the amounts available of the two sources, then the demand can be fulfilled by these sources. 5. Multi-source calculation If the required amount of a source is not available, then entirely use what is available and consider the next neighbor source (that is just higher/lower in level than the source used thus far) to satisfy the demand. In general, if Si denotes the intermediate sources entirely used so far, then the required amounts of the higher-level source Sh and lower-level source Sl for the demand Dp are given by

F Sh;Dp ¼ ½ðF Dp yDp 

X

F Si;Dp ySi Þ  ðF Dp 

X

F Si;Dp ÞySl =ðySh  ySl Þ ð7aÞ

F Sl;Dp ¼ ½ðF Dp 

X

F Si;Dp ÞySh  ðF Dp yDp 

X

F Si;Dp ySi Þ=ðySh  ySl Þ ð7bÞ

The summation in Eqs. (7) goes over all intermediate levels ySi used hitherto.

1. Matching matrix with pinch division 6. Check if all demands and sources are matched. If all demands are not met, go to step 2. 7. Validate the pre-established targets and stop.

Arrange, in order of decreasing level (i.e., purity), all sources and demands as rows and columns respectively in a matching matrix [44]. Identify cross-pinch regions to be grayed out (top-right and bottom-left cell blocks on this matching matrix as in Fig. 4) based on the pinch level already targeted.

Note that Eqs. (6) are obtained by the simultaneous solution of the following flow and bioethanol load balances:

2. Same-level sources

F Sh;Dp þ F Sl;Dp ¼ F Dp

ð8aÞ

F Sh;Dp ySh þ F Sl;Dp ySl ¼ F Dp yDp

ð8bÞ

Meet demands by sources at the same level, if such sources exist and are available in adequate amount. If the amount is inadequate, then use the entire available amount of the sources and consider two neighbor sources to meet the remaining demand as in Step 5.

Similarly, Eqs. (7) are obtained by simultaneously solving the following flow and load balances:

F Sh;Dp þ F Sl;Dp ¼ F Dp 

X

F Si;Dp

ð9aÞ

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

5

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

F Sh;Dp ySh þ F Sl;Dp ySl ¼ F Dp yDp 

X

F Si;Dp ySi

ð9bÞ

Eqs. (7) and (9) are generalizations of Eqs. (6) and (8) for the case when more than two sources are used to satisfy a demand. Eqs. (7), although different from earlier versions [42], are equivalent forms convenient for implementing on a calculator or a computer (using simply SUM and SUMPRODUCT in a spreadsheet). 2.5. Stepwise methodology Key steps of the methodology along with the relevant equations are summarized below: 1. Calculate the net system quantities D1 and D2 using Eq. (1). 2. Implement the basic UTA table (first five columns) as described in Section 2.3 using Eqs. (5a)–(5d). 3. Establish the minimum fresh bioethanol resource flowrate target FR at the two limits of Fin as the maximum value in the sixth / seventh columns of the UTA table based on Eq. (5e). 4. Identify the pinch as the purity level (in the first column of the UTA table) corresponding to the maximum FR value. 5. Establish continuous targets over the Fin range for the optimal values of FR, FE, yE, Fpr, Fr and yr as a function of Fin using Eqs. (4a)–(4f). 6. Synthesize bioethanol networks applying the NNA utilizing Eqs. (6) and (7). The following section demonstrates the application of the above stepwise methodology to target and synthesize bioethanol networks using a case study.

[13]) for the 96% purity arrow to the 70% purity mains in Fig. 3a. It must be emphasized that the data in Table 1 are identical to that of Martinez-Hernandez et al. [13], and water is not necessary as a source in the rest of the analysis. Other data [13] are as follows: fresh bioethanol is at 99.6% purity (yR = 0.996); ethanol recovery for the purification is 98% (R = 0.98); product purity is 96% (ypr = 0.96, which logically corresponds to the purity of demands D1 and D2); and purifier inlet purity is 68.22% (yin = 0.6822, which logically corresponds to the highest purity source S4 with a large flowrate going to excess/waste in the initial bioethanol network in Fig. 3). 3.2. Targeting without process constraints Implementing the UTA as described in Section 2.3 on the data in Table 1 yields Table 2. At the two limits of Fin = 0 and 26709.185 t/y (choice of value explained later), Eq. (5e) is used to generate the sixth and seventh columns of Table 2, respectively. The maximum value in these columns gives the minimum flowrate target for the fresh bioethanol as FR = 12512.396 t/y (at Fin = 0) and FR = 2199.186 t/y (at Fin = 26709.185 t/y). The corresponding level in the first column specifies the pinch at 91.52% purity at both limits of Fin. Since the pinch is held by the same point (yp = 0.9152, Mp = 1011.0016 t/y as per Table 2), Eq. (4a) gives FR0 = 12512.396 t/y and K = 0.3861297. Thus, according to Eqs. (4), the continuous targets (without process constraints) for 0 6 Fin 6 26709.185 t/y are

F R ¼ 12512:396  0:3861297 F in ;

FE

¼ 94855:396  0:3861297 F in 3. Results and discussion

yE ¼ ð12462:001  0:3845852 F in Þ=ð94855:396  0:3861297 F in Þ 3.1. Case study of bioethanol integration The case study considers a biorefinery with a processing capacity of 340,000 t/y of wheat, from which 13,600 t/y of bran is separated to produce 2460 t/y of 70% purity arabinoxylan (AX) [45]. It involves AX extraction integrated with bioethanol production, wherein ethanol streams at different purities are needed for AX precipitation and for feedstock washing [46,47]. Possible integration opportunities in terms of bioethanol pathways within the biorefinery are identified in Fig. 2 [13]. The initial bioethanol network depicted as a conventional flowsheet in Fig. 3a is used to extract the bioethanol demand and source data. Table 1 shows the data [13] in terms of ethanol purity mass fraction y and flowrate F (t/y) for seven processes comprising four demands (inlet streams) and six sources (outlet streams). The seven processes are the AX precipitation unit (PPU), ethanol washing unit (WSU), treatment unit (TMU), sieving/washing unit (SWU), rotary drying (RDY), centrifugation (CFG) and treatment/sieving/washing unit (TSW). The ethanol demands are at 96% purity (for AX precipitation PPU and ethanol washing WSU) and at 70% purity (for bran purification by treatment TMU and sieving/washing SWU). Using the data in Table 1 and Eq. (1), the net system quantities are obtained as D1 = 82,343 t/y and D2 = 0.345 t/y. The initial bioethanol flowsheet in Fig. 3a is converted in this work and conveniently represented as a matching matrix in Fig. 3b. Demands at the same purity level are combined (e.g., D1 + D2 at 96% purity, and D3 + D4 at 70% purity) to obtain a simpler compact matching matrix with fewer columns. Note that the water required for bioethanol dilution in the initial bioethanol network is not explicitly shown by Martinez-Hernandez et al. [13] and their water balance is represented as a diamond knot. However, Fig. 3 shows the water requirements explicitly and our water balance yields a flow of 20137.909 t/y (rather than 34,378 t/y

F pr ¼ 0:6964125 F in ;

F r ¼ 0:3035875 F in ;

yr ¼ 0:0449426

The targets from the above formulae evaluated at the two limits are summarized in the first two columns (without process constraints) of Table 3. Note that the fresh bioethanol is reduced from the initial 41161.358 t/y (Fig. 3a) to 12512.396 t/y (with no purification) and 2199.186 t/y (with purification) resulting in a bioethanol utility import savings of 28648.962 t/y and 38962.172 t/y, respectively. Assuming a bioethanol price of 590 £/t [45], the avoided revenue losses are estimated to be 16.90 M£/y and 22.99 M£/y, respectively. The target of 12512.396 t/y (with no purification) obtained analytically from the UTA agrees with that found graphically [13] from the composite curves and surplus diagram. The upper limit of Fin may be ascertained as follows. If Fin = FS4 = 55,633 t/y, then FR = 0 as per Eq. (5e) from the UTA; however, Eqs. (1) for FR = 0 give FE = 82,343 t/y along with a negative value of yE which is not feasible. On recognizing that yr = 0.0449426 (a constant for yin = 0.6822) and the excess/waste stream will comprise all of source S7 along with some purifier residue Pr, Eq. (1b) yields FR(0.996)  80,304(0.0249)  (FR + 82,343  80,304)(0.0449426) = 0.345 or FR = 2199.186 t/y. With yp = 0.9152 and Mp = 1011.0016 t/y, Eq. (4a) gives Fin = 26709.185 t/y. Generating the last column of Table 2 at Fin = 26709.185 t/y using Eq. (5e) and ensuring that the maximum value occurs at FR = 2199.186 t/y confirms the target. 3.3. Bioethanol network design without process constraints Bioethanol allocation networks are now synthesized to meet the targets recognizing the pinch at yp = 0.9152 using the NNA from Section 2.4. Consider the above-pinch design (top-left of the

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

6

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

Fig. 2. Bioethanol pathways in biorefinery processing wheat with arabinoxylan (AX) extraction.

matching matrix) in Fig. 4a (for the case of no purification). Demands D1 + D2 (22,567 t/y) at 0.96 purity fraction are satisfied by the nearest neighbor sources at 0.996 (Fresh or R) and 0.9152 (S2) using Eqs. (6) yielding FR,D1+D2 = 12512.396 t/y and FS2,D1+D2 = 10054.604 t/y. This above-pinch design is expectedly identical to that proposed by Martinez-Hernandez et al. [13], who did not explicitly complete the below-pinch design due to process constraints (considered here later). Consider next the below-pinch design (bottom-right of the matching matrix). Demands D3 + D4 (57,075 t/y) at 0.7 purity fraction may be satisfied by the nearest neighbor sources at 0.9152 (S2) and 0.6822 (S4) using Eqs. (6) giving FS2,D3+D4 = 4360.236 t/y and FS4,D3+D4 = 52714.764 t/y. Whatever remains of the sources (7255.16 t/y of S2, 2918.236 t/y of S4, 892 t/y of S1, 2049 t/y of S5, 1437 t/y of S6 and 80,304 t/y of S7) goes to waste/excess, which totals 94855.396 t/y at 13.14% purity and is in accordance with the targets in Table 3. This is a below-pinch design for the grassroots case (not shown as a matching matrix here). Now, consider a possible below-pinch design for the retrofit case (Fig. 4a). Keeping in mind that sources S2, S1 and S6 are fully

utilized in the existing network (Fig. 3), demands D3 + D4 may be met by S2 (21,670  10054.604 = 11615.396 t/y), S1 (892 t/y) and S6 (1437 t/y); then, Eqs. (7) give FS4,D3+D4 = 41777.597 t/y and FS7,D3+D4 = 1353.007 t/y. At this stage, what remains of the sources (13855.403 t/y of S4, 2049 t/y of S5 and 78950.993 t/y of S7) goes to waste/excess, which is in agreement with the targets (first column in Table 3). In matching matrices for retrofit cases (as in Figs. 4 and 5), cell values in bold highlight new matches and cells labeled ‘reuse’ indicate interconnections (existing pipes) from the initial network that may be possibly reused. Below each matrix, the number of matches is given as obtained by starting with the matches in the initial network and then adding the new matches and subtracting the possible-reuse interconnections. Note that water for bioethanol dilution is not required in any of the optimal networks designed by the NNA, and therefore two pure-water (0% ethanol) interconnections from the initial network in Fig. 3 are available for reuse. Having designed a network in Fig. 4a for the case of no purification, consider next the network in Fig. 4b for the case of purification (with Fin = 26709.185 t/y) based on the targets in the second

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

7

Fig. 3. Initial bioethanol network as (a) a conventional flowsheet and (b) a matching matrix.

Table 1 Bioethanol demand and source data.a,b Process (ID)

yd

ys

Fd (t/y)

Fs (t/y)

PPU (1) WSU (2) TMU (3) SWU (4) RDY (5) CFG (6) TSW (7)

0.96 0.96 0.7 0.7

0.64 0.9152

11,888 10,679 38,695 18,380

892 21,670

0.6822 0.5095 0.1509 0.0249

55,633 2049 1437 80,304

a y is ethanol purity mass fraction, F (t/y) is flowrate, subscript d is demand, subscript s is source. b D1 = 79,642  161985 = 82,343 t/y; D2 = 61616.82  61616.475 = 0.345 t/y.

(1767.203 t/y of S2). For the below-pinch design, the purifier inlet Pin is first satisfied with 26709.185 t/y of same-level source S4 at 0.6822 purity fraction. Next, demands D3 + D4 may be met by S2 (19902.797 t/y), S4 (28923.815 t/y), S1 (892 t/y), S5 (2049 t/y), S6 (1437 t/y) and Pr (3870.388 t/y) in accordance with Eqs. (7). What remains of the purifier residue Pr (4238.186 t/y) and all of lowestpurity source S7 (80,304 t/y) goes to waste/excess. The networks in Fig. 4 are classic pinch designs that allow sources to be matched with demands without restrictions, and thus do not consider any process constraints. It may be noted that practical constraints may exist in certain processes that disallow direct reuse of certain sources. The handling of process constraints during integration is discussed next. 3.4. Targeting with process constraints

column of Table 3. The inlet stream (demand) from the purifier introduces an extra column in the matching matrix, whereas the two purifier outlet streams (sources) introduce two additional rows. For the above-pinch design, demands D1 + D2 (22,567 t/y) are satisfied by 18600.61 t/y of same-level source from the purifier product Ppr at 0.96 purity fraction with the remainder by the nearest neighbor sources at 0.996 (2199.186 t/y of Fresh) and 0.9152

According to Martinez-Hernandez et al. [13], the stream leaving SWU (source S4) cannot be directly reused in the system because it contains the ethanol extractable components from the bran which are not desirable for the AX product. This practical process constraint may be accounted for by simply using the data in Table 1 with the flowrate of source S4 taken as zero (rather than 55,633 t/y).

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

8

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

Table 3 Summary of targets for yin = 0.6822 (with R = 0.98 and ypr = 0.96). Fin (t/y)

Without process constraints

FR (t/y) FE (t/y) yE Fpr (t/y) Fr (t/y) yr Bioethanol saved (t/y) Cost Savings (M£/y) Percent Savings

With process constraints

0

26709.185

0

32404.645

55,633

12512.396 94855.396 0.1314 – – – 28648.962 16.90 69.60

2199.186 84542.186 0.0259 18600.610 8108.575 0.0449 38962.172 22.99 94.66

39767.538 122110.538 0.3244 – – – 1393.82 0.82 3.39

18037.128 100380.128 0.1790 22,567 9837.645 0.0449 23124.23 13.64 56.18

2460.297 84803.297 0.0289 38743.517 16889.483 0.0449 38701.061 22.83 94.02

Fig. 4. Bioethanol allocation networks (as matching matrices) without process constraints for (a) case of no purification (Fin = 0); and (b) case of purification (Fin = 26709.185 t/y).

Implementing the UTA from Section 2.3 with this process constraint (FS4 = 0) yields Table 4. At the two limits of Fin = 0 and 55,633 t/y (where all of source S4 is sent to the purifier inlet rather than to excess/waste), Eq. (5e) is utilized to obtain the sixth and seventh columns of Table 4, respectively. The maximum value in these columns gives the minimum fresh bioethanol flowrate target as FR = 39767.538 t/y (at Fin = 0) and FR = 2460.297 t/y (at Fin = 55,633 t/y). The corresponding level in the first column identifies the pinch at 2.49% purity at both limits of Fin. The pinch is held by the same point (yp = 0.0249, Mp = 38618.2566 t/y as per Table 4); so, Eq. (4a) gives FR0 = 39767.538 t/y and K = 0.6705955. As per Eqs. (4), the continuous targets (with process constraints) for 0 6 Fin 6 55,633 t/y are

value is calculated from Fin = Fpr/0.6964125 so that the purified product flowrate Fpr exactly meets 22,567 t/y for demands D1 + D2 at 96% purity. With the process constraints, the fresh bioethanol is reduced from the initial 41161.358 t/y to 39767.538 t/y (without purification), 18037.128 t/y (with purification of S4 partially) and 2460.297 t/y (with purification of S4 totally) giving a bioethanol utility import savings of 1393.82 t/y, 23124.23 t/y and 38701.061 t/y, respectively (translating to cost saving estimates of 0.82 M£/y, 13.64 M£/y and 22.83 M£/y, respectively). The target of 2460.297 t/y (with purification of total S4) obtained analytically from the UTA is consistent with the value of 2459 t/y found graphically [13].

F R ¼ 39767:538  0:6705955 F in ; F E ¼ 122110:538  0:6705955 F in yE ¼ ð39608:123  0:6679132 F in Þ=ð122110:538  0:6705955 F in Þ F pr ¼ 0:6964125 F in ;

F r ¼ 0:3035875 F in ;

yr ¼ 0:0449426

Targets from the above formulae calculated at the two limits and an additional intermediate value (Fin = 32404.645 t/y) are summarized in the last three columns of Table 3. The intermediate

3.5. Bioethanol network design with process constraints Fig. 5 shows bioethanol allocation networks designed by the NNA to meet the targets with process constraints (last three columns of Table 3). Since the pinch is now at yp = 0.0249, there are no cross-pinch regions in the matching matrix. Cells labeled X indicate forbidden matches due to process constraints.

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

9

Fig. 5. Bioethanol allocation networks (as matching matrices) with process constraints for (a) no purification (Fin = 0); (b) purification of S4 partially (Fin = 32404.645 t/y); (c) purification of S4 totally (Fin = 55,633 t/y); and (d) purification over entire range (0 6 Fin 6 55,633).

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

10

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

(a)

(b)

Fig. 6. Final bioethanol allocation networks (as conventional flowsheets) with process constraints for (a) no purification; and (b) purification (as superstructure with values at Fin = 55,633 t/y).

In Fig. 5a (for the case with no purification), source S4 cannot be used to meet any demands and must necessarily go to excess/ waste. Demands D1 + D2 are fulfilled in Fig. 5a exactly the same

way as in Fig. 4a using Eqs. (6). Next, demands D3 + D4 may be met by fresh bioethanol (27255.142 t/y), S2 (11615.396 t/y), S1 (892 t/y), S5 (2049 t/y), S6 (1437 t/y) and S7 (13826.462 t/y) as

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

11

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

3.6. Graphical understanding of targeting with purification The approach thus far has been completely analytical without use of any graphical procedures. In this subsection, the graphical basis underlying the targeting methodology is briefly explained to better understand its rationale. The limiting composite curves (LCCs) obtained on plotting the basic UTA data (specifically, C  1  y using the first column vs. M from the fifth column of Tables 2 and 4) are shown in Fig. 7a (without process constraints) and Fig. 7b (with process constraints). Note that the LCC here is classically plotted in terms of impurity fraction C to maintain consistency with its analog of contaminant concentration in water networks [23,27]. For the case without purifier, the target profile (shown dashed in Fig. 7a and b) simply consists of a single fresh resource line [with slope = 1/FR throughout]. Targeting graphically for the case without purifier involves rotating the fresh resource line with the pivot on the vertical axis at (1  yR) until it just touches the LCC. This rotated line (which can never be above the LCC) with maximum slope (of 1/FR) specifies the minimum fresh bioethanol resource flowrate target FR. The point where the target profile touches the LCC defines the pinch. This graphical procedure is similar in principle to that suggested for water networks [48,49], hydrogen networks [23], and carbon emission networks [32,33].

1

(a)

0.9

1 − Purity Fraction

0.8 0.7 0.6 0.5

Limiting Composite

0.4 0.3 0.2

Target Profile Without Purifier

Pinch

0.1

1/(FR + Fpr)

1 – ypr

1/FR

1 – yR

0

500

1000

1500

Target Profile With Purifier

2000

2500

3000

Mass Load (t/y) 1

Pinch

(b)

0.9 0.8

1 − Purity Fraction

per Eqs. (7). All of source S4 (55,633 t/y) along with the remainder of source S7 (66477.538 t/y) goes to waste/excess. In Fig. 5b and c (for the two cases with purification at Fin = 32404.645 t/y and Fin = 55,633 t/y), source S4 cannot be used to meet any demands and must necessarily go to the purifier inlet (being a same-level source at 0.6822 purity fraction) with the remainder, if any, to excess/waste. The purifier residue Pr also cannot be used to satisfy any demands and must of necessity go to excess/waste. Demands D1 + D2 (22,567 t/y) are met by same-level source at 0.96 purity fraction from the purifier product Ppr. As before, demands D3 + D4 are fulfilled by appropriate sources in accordance with Eqs. (7). Finally, all of the purifier residue Pr along with the remainder of source S7 (and S4, if any) goes to waste/ excess. The three networks in Fig. 5a–c are for specific Fin values; however, they may be unified and represented as in Fig. 5d through a network superstructure, wherein the design is valid continuously for the entire range (0 6 Fin 6 55,633). The values in the matching matrix are given in the box below it by the formulae in terms of Fin based on Eqs. (6) and (7); thus, Fig. 5d yields Figs. 5a, 5b or 5c on setting Fin to 0, 32404.645 or 55,633 t/y, respectively. Fig. 5a saves 0.82 M£/y through reducing the fresh bioethanol use by 1393.82 t/y and calls for few modifications with no new purifier. In contrast, Fig. 5b and c save 13.64 M£/y and 22.83 M£/ y through large reductions in the fresh bioethanol use of 23124.23 t/y and 38701.061 t/y respectively, but entail relatively more modifications with the addition of a new purifier unit. Based on the matching matrices in Fig. 5, the final bioethanol networks are shown in their more natural flowsheet representation in Fig. 6a (without purification) and Fig. 6b (with purification and values specifically from Fig. 5c). Bold arrows highlight the new matches required in the retrofit. Dotted arrows in Fig. 6b show the additional three matches (corresponding to FR,D1+D2, FS2,D1+D2 and FS4,E) from the superstructure in Fig. 5d. The design in Fig. 6a without a purifier, which potentially yields a profit increase of 0.82 M£/ y, is not reported by Martinez-Hernandez et al. [13]; on the other hand, the design in Fig. 6b with a purifier (and values from Fig. 5c for Fin = 55,633 t/y) is essentially identical to that of Martinez-Hernandez et al. [13], who estimated the annualized capital cost for the additional purifier as 0.11 M£/y resulting in a net profit increase of 22.72 M£/y.

0.7

Limiting Composite

0.6 0.5 Target Profile With Purifier

0.4 0.3

1/(FR + Fpr)

0.2

Target Profile Without Purifier

0.1

1/FR

1 – ypr 1 – yR

0

5000

10000

15000

20000

25000

30000

35000

40000

Mass Load (t/y) Fig. 7. Target profiles to match limiting composite curve: (a) without process constraints; and (b) with process constraints.

For the case with purifier, the target profile (shown bold in Fig. 7a and b) comprises the fresh resource line segment [with slope = 1/FR for ypr 6 y 6 yR] as well as the composite line segment of the fresh resource and purified product [with slope = 1/(FR + Fpr) for y 6 ypr]. The general equations for the target profile are given by

M ¼ F R ðyR  yÞ for ypr 6 y 6 yR

ð10aÞ

M ¼ F R ðyR  yÞ þ F pr ðypr  yÞ for y 6 ypr

ð10bÞ

When there is no purifier (Fpr = 0), Eqs. (10) reduce to a single equation of the form M = FR(yR  y) for all y 6 yR. It may be noted that the shape of the LCC in Fig. 7a (without process constraints on including source S4) is dramatically different from that in Fig. 7b (with process constraints on excluding source S4). Accordingly, the pinch occurs at Cp = 0.0848 (91.52% purity) in Fig. 7a and at Cp = 0.9751 (2.49% purity) in Fig. 7b. Note that the fresh resource line segment (small bold portion for 0.96 6 y 6 0.996 or 0.004 6 C 6 0.04) is visible in Fig. 7a, but is not clearly visible in Fig. 7b being virtually coincident with the vertical axis. 4. Conclusions A systematic straightforward methodology, which effectively depends on the UTA for targeting and the NNA for network design, has been developed for efficient bioethanol utilization during biorefinery integration with the following significant advantages. First, the methodology is analytical and does not require any iterative or

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066

12

A.U. Shenoy, U.V. Shenoy / Energy Conversion and Management xxx (2014) xxx–xxx

graphical procedures. Second, it may be used to continuously target and synthesize bioethanol networks with purification over the entire Fin range. Third, the design without purification simply turns out to be a special case at the lower limit of Fin = 0. Fourth, the method allows process constraints to be considered with and without purification. Fifth, the approach is based on six fundamental equations involving eight design variables, and the solution strategy on specifying two variables (related to the purifier inlet stream) allows the remaining six variables to be readily evaluated in any sequential order from simple formulae. Finally, the methodology has been specifically applied here to bioethanol; however, the approach is generic enough to be extended to other analogous product-based biorefineries. Note that there are typically numerous possible networks that satisfy the target even at a particular value of Fin; however, they were not observed in this case study because a retrofit case was being examined with demands essentially at only two levels. For a grassroots case with several demands at different levels, many optimum networks may be designed using the NNA by merely changing the order in which the demands are satisfied. Thus, the NNA along with the matching matrix representation provides the engineer an effective design tool in practice to methodically synthesize and evolve many promising networks [34]. A network may be finally selected for implementation incorporating practical considerations such as layout, safety, controllability, operability, flexibility, complexity and cost. Future work may be directed toward extending the above methodology to more rigorous cost-benefit analysis and optimization. References [1] Kemp IC. Pinch analysis and process integration. Amsterdam: ButterworthHeinemann; 2007. [2] Foo DCY. Process integration for resource conservation. Boca Raton: CRC Press; 2012. [3] Klemes JJ, Varbanov PS, Kravanja Z. Recent developments in process integration. Chem Eng Res Des 2013;91(10):2037–53. [4] Klemes J, Friedler F, Bulatov I, Varbanov P. Sustainability in the process industry: integration and optimization. New York: McGraw Hill; 2010. [5] El-Halwagi MM. Sustainable design through process integration. Amsterdam: Butterworth-Heinemann; 2011. [6] Fujimoto S, Yanagida T, Nakaiwa M, Tatsumi H, Minowa T. Pinch analysis for bioethanol production process from lignocellulosic biomass. Appl Therm Eng 2011;31(16):3332–6. [7] Kravanja P, Modarresi A, Friedl A. Heat integration of biochemical ethanol production from straw – a case study. Appl Energy 2013;102:32–43. [8] Ng DKS. Automated targeting for the synthesis of an integrated biorefinery. Chem Eng J 2010;162:67–74. [9] Ponce-Ortega JM, Pham V, El-Halwagi MM, El-Baz AA. A disjunctive programming formulation for the optimal design of biorefinery configurations. Ing Eng Chem Res 2012;51(8):3381–400. [10] Pham V, El-Halwagi MM. Process synthesis and optimization of biorefinery configurations. AIChE J 2012;58(4):1212–21. [11] Dias MOS, Modesto M, Ensinas AV, Nebra SA, Filho RM, Rossell CEV. Improving bioethanol production from sugarcane: evaluation of distillation, thermal integration and cogeneration systems. Energy 2011;36(6):3691–703. [12] Modarresi A, Kravanja P, Friedl A. Pinch and exergy analysis of lignocellulosic ethanol, biomethane, heat and power production from straw. Appl Therm Eng 2012;43:20–8. [13] Martinez-Hernandez E, Sadhukhan J, Campbell GM. Integration of bioethanol as an in-process material in biorefineries using mass pinch analysis. Appl Energy 2013;104:517–26. [14] Costa ALH, Queiroz EM. An extension of the problem table algorithm for multiple utilities targeting. Energy Convers Manage 2009;50(4):1124–8. [15] Castier M. Rigorous multiple utility targeting in heat exchanger networks. Energy Convers Manage 2012;59(7):74–85. [16] Linnhoff B. Pinch analysis: a state-of-the-art overview. Chem Eng Res Des 1993;71(5):503–22. [17] Cherubini F. The biorefinery concept: using biomass instead of oil for producing energy and chemicals. Energy Convers Manag 2010;51(7):1412–21. [18] Demirbas MF, Balat M, Balat H. Potential contribution of biomass to the sustainable energy development. Energy Convers Manage 2009;50(7): 1746–60.

[19] Demirbas A. Biofuels sources, biofuel policy, biofuel economy and global biofuel projections. Energy Convers Manage 2008;49(8):2106–16. [20] Demirbas A. Biofuels securing the planet’s future energy needs. Energy Convers Manage 2009;50(9):2239–49. [21] Alves JJ, Towler GP. Analysis of refinery hydrogen distribution systems. Ind Eng Chem Res 2002;41:5759–69. [22] Hallale N. A new graphical targeting method for water minimization. Adv Environ Res 2002;6:377–90. [23] Agrawal V, Shenoy UV. Unified conceptual approach to targeting and design of water and hydrogen networks. AIChE J 2006;52(3):1071–82. [24] Stuart PR, El-Halwagi MM. Integrated biorefineries: design, analysis and optimization. Boca Raton: CRC Press; 2013. [25] Linnhoff B, Townsend DW, Boland D, Hewitt GF, Thomas BEA, Guy AR, et al. User guide on process integration for the efficient use of energy. Rugby: IChemE; 1982. [26] Shenoy UV. Heat exchanger network synthesis: process optimization by energy and resource analysis. Houston: Gulf Publishing; 1995. [27] Shenoy UV. Unified targeting algorithm for diverse process integration problems of resource conservation networks. Chem Eng Res Des 2011; 89(12):2686–705. [28] Prakash R, Shenoy UV. Targeting and design of water networks for fixed flowrate and fixed contaminant load operations. Chem Eng Sci 2005;60(1): 255–68. [29] Shenoy UV. A unified targeting algorithm for diverse process integration problems. In: Klemes JJ, editor. Handbook of process integration: minimization of energy and water use waste and emissions. Cambridge: Woodhead Publishing/Elsevier; 2013 [chapter 18]. [30] Shukla SK, Kumar V, Yeom IT, Bansal MC. Recycling of bleach plant effluent of an Indian paper mill using water cascade analysis technique. Clean Technol Environ Policy 2012;14(4):677–85. [31] Shukla SK, Kumar V, Chakradhar B, Kim T, Bansal MC. Designing plant scale process integration for water management in an Indian paper mill. J Environ Manage 2013;128:602–14. [32] Shenoy UV. Targeting and design of energy allocation networks for carbon emission reduction. Chem Eng Sci 2010;65(23):6155–68. [33] Shenoy AU, Shenoy UV. Targeting and design of energy allocation networks with carbon capture and storage. Chem Eng Sci 2012;68(1):313–27. [34] Das AK, Shenoy UV, Bandyopadhyay S. Evolution of resource allocation networks. Ind Eng Chem Res 2009;48(15):7152–67. [35] Tian JR, Zhou PJ, Lv B. A process integration approach to industrial water conservation: a case study for a Chinese steel plant. J Environ Manage 2008; 86(4):682–7. [36] Deng C, Feng X. Optimal water network with zero wastewater discharge in an alumina plant. WSEAS Trans Environ Dev 2009;2(5):146–56. [37] Foo DCY, Ng DKS. Process integration for cleaner process design. In: Klemes JJ, editor. Handbook of process integration: minimization of energy and water use, waste and emissions. Cambridge: Woodhead Publishing/Elsevier; 2013 [chapter 14]. [38] Tan RR, Foo DCY. Pinch analysis for sustainable energy planning using diverse quality measures. In: Klemes JJ, editor. Handbook of process integration: minimization of energy and water use, waste and emissions. Cambridge: Woodhead Publishing/Elsevier; 2013 [chapter 17]. [39] Saw SY, Lee L, Lim MH, Foo DCY, Chew IML, Tan RR, et al. An extended graphical targeting technique for direct reuse/recycle in concentration and property-based resource conservation networks. Clean Technol Environ Policy 2011;13:347–57. [40] Bandyopadhyay S, Mishra M, Shenoy UV. Energy-based targets for multiplefeed distillation columns. AIChE J 2004;50(3):1837–53. [41] Hallale N, Liu F. Refinery hydrogen management for clean fuels production. Adv Environ Res 2001;6:81–98. [42] Shenoy UV. Enhanced nearest neighbors algorithm for design of water networks. Chem Eng Sci 2012;84:197–206. [43] Shenoy AU, Shenoy UV. Targeting and design of CWNs (cooling water networks). Energy 2013;55:1033–43. [44] Prakash R, Shenoy UV. Design and evolution of water networks by source shifts. Chem Eng Sci 2005;60:2089–93. [45] Sadhukhan J, Mustafa MA, Misailidis N, Mateos-Salvador F, Du C, Campbell GM. Value analysis tool for feasibility studies of biorefineries integrated with value added production. Chem Eng Sci 2008;63:503–19. [46] Du C, Campbell GM, Misailidis N, Mateos-Salvador F, Sadhukhan J, Mustafa M, et al. Evaluating the feasibility of commercial arabinoxylan production in the context of a wheat biorefinery principally producing ethanol. Part 1. Experimental studies of arabinoxylan extraction from wheat bran. Chem Eng Res Des 2009;87:1232–8. [47] Misailidis N, Campbell GM, Du C, Sadhukhan J, Mustafa M, Mateos-Salvador F, et al. Evaluating the feasibility of commercial arabinoxylan production in the context of a wheat biorefinery principally producing ethanol. Part 2. Process simulation and economic analysis. Chem Eng Res Des 2009;87:1239–50. [48] Wang YP, Smith R. Wastewater minimisation. Chem Eng Sci 1994;49(7): 981–1006. [49] Feng X, Huang L, Zhang X, Liu Y. Water system integration of a brewhouse. Energy Convers Manage 2009;50(2):354–9.

Please cite this article in press as: Shenoy AU, Shenoy UV. Designing optimal bioethanol networks with purification for integrated biorefineries. Energy Convers Manage (2014), http://dx.doi.org/10.1016/j.enconman.2014.02.066