Heat-pump assisted distillation versus double-effect distillation for bioethanol recovery followed by pressure swing adsorption for bioethanol dehydration

Accepted Manuscript Heat-Pump Assisted Distillation versus Double-Effect Distillation for Bioethanol Recovery followed by Pressure Swing Adsorption for Bioethanol Dehydration Ashish Singh, Sergio da Cunha, G.P. Rangaiah PII: DOI: Reference:

S1383-5866(18)31895-1 https://doi.org/10.1016/j.seppur.2018.08.043 SEPPUR 14864

To appear in:

Separation and Purification Technology

Received Date: Revised Date: Accepted Date:

31 May 2018 8 August 2018 23 August 2018

Please cite this article as: A. Singh, S. da Cunha, G.P. Rangaiah, Heat-Pump Assisted Distillation versus DoubleEffect Distillation for Bioethanol Recovery followed by Pressure Swing Adsorption for Bioethanol Dehydration, Separation and Purification Technology (2018), doi: https://doi.org/10.1016/j.seppur.2018.08.043

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Heat-Pump Assisted Distillation versus Double-Effect Distillation for Bioethanol Recovery followed by Pressure Swing Adsorption for Bioethanol Dehydration Ashish Singh, Sergio da Cunha, G.P. Rangaiah* Department of Chemical and Biomolecular Engineering National University of Singapore, Singapore 117585 *Corresponding author: [email protected] Abstract Bioethanol recovery (i.e., pre-concentration from 5-10 wt.% ethanol to near azeotrope composition) from the fermentation broth requires significantly more energy compared to bioethanol dehydration from near azeotrope composition to > 99.5 wt.% ethanol. The present article proposes and evaluates two different alternatives, namely, double-effect distillation (DED) and heat-pump assisted distillation (HPAD) as opposed to simple distillation for bioethanol recovery followed by dehydration using pressure-swing adsorption (PSA). Preliminary design reveals that HPAD-PSA process outperforms DED-PSA process resulting in 19% savings in total annual cost. Multi-objective optimization is hence performed for HPAD-PSA process, and the obtained Pareto-optimal solutions for minimizing fixed capital investment and annual operating cost are presented and discussed. Compared to the simple distillation followed by PSA process, HPAD-PSA process reduces specific energy consumption (SEC) by 38% to 3.2 MJ-fuel/kg of bioethanol. Further, compared with other, recent bioethanol separation processes, HPAD-PSA process results in 35-64% lesser SEC. Keywords: Bioethanol Separation; Double-effect

Distillation; Heat-Pump Assisted

Distillation; Pressure-Swing Adsorption; Multi-Objective Optimization 1. Introduction The past decade has witnessed a flurry of interest over bioethanol as a renewable energy source for mitigation of greenhouse gas emissions especially from the transport sector. Today, several technologies exist for bioethanol recovery and dehydration to produce > 99.5 wt.% bioethanol from a dilute range of 5-20 wt.% ethanol and remaining water [1]. Notable among them include vapor compression distillation followed by salt-based extractive distillation [2], heat-pump (HP) assisted single extractive dividing wall column (E-DWC) [3], E-DWC [4,5], double effect distillation (DED) followed by EDWC [6], EDWC with three 1

dividing walls [7], and distillation followed by vapor permeation (D-VP) [8]. Although DWC technologies are given a green signal in industry practice [9,10], to the best of authors’ knowledge, the applicability of these technologies is yet to be seen in the bioethanol industry. In recent years, hybrid separation technologies involving membranes have gained attention; however, they also lack industry acceptance for large bioethanol production volumes, primarily due to high cost of membrane, membrane fouling concerns, shorter lifetime and low separation factor of polymeric membranes. Amongst the hybrid separation technologies, distillation followed by pressure swing adsorption (D-PSA) for bioethanol recovery and dehydration is widely adopted in industries in the United States, which is one of the largest producers of bioethanol [11–13]. In this context, Loy et al. [5] studied D-PSA process and found the cost of manufacturing (COM) of bioethanol to be the lowest in comparison to two commonly studied technologies, namely, distillation followed by extractive distillation (ED) using ethylene glycol (16% expensive) and EDWC using ethylene glycol (49% expensive, mainly due to solvent make-up cost). Further, D-PSA was found to be economical in terms of COM over the D-VP process (45% expensive, mainly due to higher membrane costs) proposed by Singh and Rangaiah [8]. Note that fixed capital investment (FCI) used for evaluating COM for the above-stated technologies are based on total module cost (TMC). From the point of energy requirement, D-VP process was found to be 14% less energy intensive as compared to D-PSA. Motivated by this, we identified two possible technologies, namely, DED and HP-assisted distillation (HPAD) to reduce the energy requirement in the pre-concentration section. Thus, the present study proposes two processes: DED followed by PSA (DED-PSA) and HPAD followed by PSA (HPAD-PSA) for bioethanol recovery and dehydration using the feed conditions in Loy et al. [5]. Novelty and significance of the present study are the comparative study of process development and optimization of DED-PSA and HPAD-PSA process for the first time. Moreover, there are fewer studies on PSA-based processes for bioethanol separation [1]. Two conflicting objectives, namely, FCI and annual operating cost (AOC) are simultaneously minimized by multi-objective optimization (MOO) using the Excel-based multi-objective differential evolution program [14]. For this, each of the proposed separation processes simulated in Aspen HYSYS is interfaced with the Excel-based MOO program via Visual Basic for Applications (VBA), using HYSYS V9.0 type library. For conflicting objectives, MOO provides a set of solutions known as Pareto-optimal or non-dominated solutions, which are equally good from the perspective of the objectives employed. Based on

2

these solutions, optimal values of total annualized cost (TAC) are analyzed for a given payback period. The rest of this article is organized as follows. Section 2 describes the development of the two hybrid processes: DED-PSA and HPAD-PSA for bioethanol recovery and dehydration. Section 3 deals with equipment sizing and economic evaluation. Section 4 discusses the preliminary design and economics of the proposed processes. Section 5 formulates an MOO problem for the economical hybrid process (HPAD-PSA) found in Section 4. Section 6 presents and discusses MOO results obtained for HPAD-PSA process. Lastly, main findings of this study are summarized in the Conclusions section. 2. Hybrid processes for bioethanol recovery and dehydration - development and simulation D-PSA process studied by Loy et al. [5] for producing 200,000 m3/year of ethanol is described here to establish the base for the proposed hybrid processes. In this process, whose flowsheet is given in Figure S1 in the Supplementary Material, a ternary feed of 10 wt.% ethanol, 89.9 wt.% water and 0.1 wt.% CO2 is preheated in a series of heat exchangers before sending to the distillation column (DC1). The overhead vapor leaving this column mostly contains CO2; it is cooled with cooling water to 40°C and then sent to the degasser (similar to a vapor-liquid flash/separator). The degasser produces two outlet streams; CO 2 rich-vapor stream from the top is purged, and the liquid stream is pumped and recycled to 1st stage (from the top) of DC1, thereby minimizing ethanol loss from the process. The side draw (vapor stream having ~55 wt.% ethanol) from DC1 enters another distillation column (DC2) for further concentration of ethanol. The overhead stream (having ~92.6 wt.% ethanol) from the second distillation column is compressed to 200 kPa before it enters the 2-column PSA system (using zeolites) for dehydration to produce 99.8 wt.% bioethanol. For details pertaining to the functioning of PSA system, see Loy et al. [5]. The desorbed stream from the PSA system is compressed (2-stage compression with inter-cooling) and is recycled to 1st stage (from the top) of the second distillation column. Loy et al. [5] included suitable heat integration in D-PSA process to minimize heating/cooling utilities required. 2.1 Double effect distillation followed by pressure swing adsorption process The first distillation column in the pre-concentration section of the D-PSA process is replaced by two columns in parallel (see Figure 1), operating at different pressures. The low pressure column (LPC) operates at 110 kPa, whereas the high pressure column (HPC) 3

operates at 890 kPa. The ternary feed, having composition same as in Loy et al. [5], is divided into two streams by a splitter (FS-101). One of them enters the LPC after preheating, while the other is sent to HPC after preheating. The overhead vapor streams leaving LPC and HPC, containing mostly CO2, are mixed (MIX-101) and processed as in the D-PSA process. The liquid stream from the degasser is recycled to the 1st stage of LPC alone, owing to its relatively small flowrate. The side draw from LPC and the depressurized side draw from the HPC are mixed (MIX102) before sending to another distillation column (DC2) for further concentration of ethanol. The overhead stream from DC2 is compressed to 221 kPa and cooled before it enters the 2column PSA system (using zeolites). Note that the PSA column shown in Figure 1 is an equivalent representation of 2-column PSA system. After adsorption in the PSA system, the dehydrated ethanol (99.8 wt.%) undergoes vapor recompression for use as a heat source in DC2 reboiler. After the dehydrated ethanol vapor is used for vaporizing in HE-114, it is used for preheating (in HE-102) the feed entering the LPC. Finally, it is cooled (in HE-115) to the desired temperature of 40°C for storage. The bottom streams of all distillation columns contain mostly water (99.99 wt.%). Table 1 lists the data of key process streams in Figure 1. The purge stream (Stream 41) of the PSA system contains substantial amount of ethanol. Hence, it is sent to a series of three compressors with cooling in-between, for increasing its pressure for recycling to the 15th stage (from top) of DC2. Three compressors are required for the following reasons. First, this superheated stream enters the compressor at 101.8°C. Second, the stream temperature at the outlet of each compressor should not rise above 195°C. This temperature limit is within the range 167-204°C for maximum compressor exit temperatures given by Turton et al. [15]. Third, each inter-cooling heat exchanger has a pressure drop of 21 kPa on the vapor side [16]. Alternatively, a single compressor with three stages (and inter-stage cooling) can be used to compress stream 41 in Figure 1. This will result in marginal savings in the capital cost. Heat integration is implemented in DED-PSA process using the Pinch Analysis Spreadsheet in [17]. A minimum temperature approach (∆T min) of 10°C is chosen for heat integration between possible streams except for heat integration between LPC reboiler and HPC condenser where ∆Tmin is assumed to be 5°C. Note that, for clarity in Figure 1, two separate heat exchangers are shown for depicting heat integration. For example, HE-101a and HE101b correspond to one single heat exchanger (HE-101), where Stream 57 is used for

4

preheating part of LPC inlet stream (Stream 4). Following are the heat integrations in DEDPSA process (Figure 1). 1. Feed preheating before entering LPC: preheating of Streams 4 and 5 using Streams 57 and 49 in HE-101 and HE-102 respectively, in parallel mode. 2. Feed preheating before entering HPC: preheating Stream 8 using Stream 13 in HE103. 3. Double effect operation: Stream 21 with Stream 18 in HE-107, i.e. heat exchange between HPC condenser and LPC reboiler. 4. Vapor recompression of Stream 47 to provide heat to DC2 reboiler (HE-114) 5. Use of sensible heat of cooling Streams 12 and 16 to partially vaporize the bottoms of LPC and DC2 in HE-105 and HE-106, respectively. See Appendix A for more details on the heat exchanger network (HEN).

Figure 1: Preliminary design of DED-PSA process with LPC at 110 kPa and HPC at 890 kPa; stream data can be found in the Supplementary Material (Table S1), and numbers in the column refer to ideal stages. The DED-PSA process illustrated in Figure 1 has extensive energy integration. Therefore, this process has more interactions among the units, which may cause control issues. If the 5

control of the DED-PSA process in Figure 1 is found to be infeasible, minor changes can be done in the original design. In Figure 1, all the thermal energy to LPC reboiler is provided by process streams. To provide flexibility for control, HE-105 duty can be reduced slightly and a trim reboiler (with small duty supplied by steam) can be installed for LPC. The steam flowrate to the trim reboiler can be manipulated for process control purposes. Installation of trim condenser and/or reboiler is a common practice to enhance controllability of DED configurations [18]. Although this requires the use of some steam for LPC trim reboiler, feed split ratio in the DED-PSA process can be changed to minimize hot utility, which includes the reboiler duty of HPC (HE-104) and steam supplied to LPC trim reboiler. 2.2 Heat-pump assisted distillation followed by pressure swing adsorption process In HPAD-PSA configuration, the second distillation column of D-PSA process studied by Loy et al. [5] is modified to accommodate vapor recompression technology. In [5], the overhead vapor from the second distillation column is compressed to increase the pressure of the stream for adsorption in the 2-column PSA system; however, the compression is not for heat integration. In order to utilize the heat of compression, we propose two scenarios using vapor recompression technology as follows. In the preliminary design of the Scenario 1 (Figure 2), the overhead stream (Stream 25) from DC2 is initially compressed (pressure ratio ~1.8) to an intermediate pressure of 221 kPa and is split into two streams (Streams 27 and 33). Stream 33 is slightly cooled before it enters the 2-column PSA system. The other stream (Stream 27) is further compressed (pressure ratio ~2.1) to a sufficiently higher pressure of 474 kPa for heat integration. After adsorption of Stream 34 in the PSA system, dehydrated ethanol (99.8 wt.%) undergoes vapor recompression for use as a heat source in DC1 reboiler (HE-104). Then, it is cooled to 40°C for storage. The purge stream (Stream 35) of the PSA system contains substantial amount of ethanol, and hence it is sent to a series of three compressors with cooling in-between, for recycling to the 12th stage (from top) of DC2. Reasons for using three compressors are given in Section 2.1. The bottom streams of all distillation columns contain mostly water (99.9 wt.%). Table 1 lists the data of key process streams in Figure 2.

6

Figure 2: Optimal design of HPAD-PSA process for the chosen optimal solution with lowest TAC (described in Section 6.1); DC1 and DC2 are at 125 kPa and 120 kPa respectively, numbers in the columns refer to ideal stages, and stream data can be found in the Supplementary Material (Table S2). As in Section 2.1, heat integration using pinch analysis spreadsheet is implemented for HPAD-PSA process with ∆Tmin of 10°C for all exchangers including reboilers and condensers. Following are the heat integrations for HPAD-PSA process in Figure 2. 1. Feed preheating before entering DC1: preheating Stream 2 using Stream 23 in HE101 followed by preheating Stream 3 using sensible heat of Stream 30 in HE-102. 2. Sensible and latent heat of Stream 28 (as a result of vapor recompression of Stream 27) and sensible and latent heat of Stream 42 (as a result of vapor recompression of Stream 41) to provide DC1 reboiler duty, in HE-103 and HE-104 respectively. 3. Latent and sensible heat of Stream 29 (as a result of vapor recompression of Stream 27 to provide part of DC2 reboiler duty (HE-107). See Appendix A for more details on the HEN. Similar to the DED-PSA process, energy integration in the HPAD-PSA configuration is extensive and can lead to problems in process control and operation. However, minor changes in the process design can improve its controllability, if required. For the HPAD7

PSA process in Figure 2, one possible modification is based on the work by Patrascu et al. [19], and it consists of two key points. Firstly, small quantity of external steam is introuced into HE-103. In order to implement this, area for heat transfer between streams 28 and 5 should be slightly smaller than that for the flowsheet in Figure 2, and area for heat transfer between steam and stream 5 should be added, both within HE-103. This is possible by installing separate tubes for stream 28 and steam. This way, the small deficit of reboiler duty for DC1 will be supplied by utility/steam. Secondly, since stream 29 vapor fraction increases because of the previous change, duty of HE-107 can be increased and duty of HE-108 (using steam) can be decreased. Note that HE-107 and HE-108 together provide the reboiler duty of DC2. In this way, some steam is used for reboiler duty of both DC1 and DC2, and this increases flexibility for control of both these columns. An alternate modification of the process in Figure 2, not requiring external steam supply for DC1, is to split stream 28 into three sub-streams to be fed to HE-102, HE-103 and HE-107. This modification is motivated by the work of Luyben [20]. In our process, this increases the ‘degrees of freedom’ for control, as the split fraction of stream 28 to HE-103 can be manipulated, for example, to maintain the temperature of a given stage in DC1 (i.e., indirect composition control), and split fraction to HE-102 can be manipulated to maintain the temperature of stream 4 entering DC1. Either this or the previous modification may increase the capital and utility costs of the process marginally. In Scenario 2 (figure not shown for brevity), the overhead stream from DC2 is compressed to 474 kPa using only one compressor (instead of two in the first scenario). This pressurized stream is then split into two streams. One of them is depressurized, cooled and then sent to the PSA unit. The second stream is used for heat integration, and then recycled back to DC2 (similar to Stream 27 in Figure 2). In both the scenarios discussed, the purged stream (Stream 35 at ~ 25 kPa) from the PSA system is compressed (3-stage compression with inter-stage cooling) to ~125 kPa and is recycled to the 12th stage (from the top) of DC2. Upon evaluation (described in Section 4.2), it was observed that savings in electricity cost for compression in Scenario 1 offsets the lower investment cost of one compressor in Scenario 2, resulting in Scenario 1 to be the better configuration for HPAD-PSA process.

8

Table 1: Mass flowrate and compositions of key process streams in Figures 1 and 2. Process

DED-PSA

HPAD-PSA

Stream number

Mass flowrate (kg/h)

1 (Feed) 31 (CO2 purge) 51 (Product)* 59 (Waste water) 1 (Feed) 14 (CO2 purge) 24 (Waste water) 44 (Product)*

1.93×105 238 1.93×104 1.73×105 1.93×105 225 1.73×105 1.93×104

Mass fractions (Ethanol, Water, CO2) (0.100, 0.899, 0.001) (0.170, 0.021, 0.809) (0.998, 0.002, 0.000) (0.0001, 0.9999, 0.000) (0.100, 0.899, 0.001) (0.133, 0.014, 0.854) (0.0001, 0.9999, 0.000) (0.998, 0.002, 0.000)

* Bioethanol recovery in both the processes is 99.7%. 2.3 Thermodynamic property package and validation Choosing a proper thermodynamic model is usually the first step in process simulation. NRTL-Redlich-Kwong equation of state (RK-EoS) using the parameters listed in Table S3 in the Supplementary Material predicts well the VLE data of ethanol and water binary mixture [5]. In the present work too, NRTL model is used to model liquid phase fugacities of water and ethanol. CO2 is treated as a Henry compound in Aspen HYSYS, since the temperature profile along each of the distillation columns is higher than its critical temperature (31.1°C). Vapor-liquid equilibrium equations and binary interaction parameters can be found in the Supplementary Material. Figure 3 shows the plot of (a) validation of solubility of CO2 in water and (b) validation of solubility of CO2 in ethanol. The experimental data are taken from Dalmolin et al. [21]. Even though this figure shows good agreement for pure solvent + CO 2, the model overestimates by a factor of up to 3 when both water and ethanol are present in the liquid phase. Therefore, the process developed here can be seen as slightly overdesigned, since separation of CO2 from the liquid mixture in simulation is harder than that in the actual process. Furthermore, since concentration of CO2 in the liquid phase is always smaller than 0.03 for both experimental data and model predictions, discrepancies in liquid fugacity of solvents [22].

9

should not affect the

Figure 3: Plot of experimental and model predictions from Aspen HYSYS for validation of solubility of CO2 in water in the liquid phase (plot a) and for validation of solubility of CO2 in ethanol in the liquid phase (plot b). 3. Equipment sizing and Economic evaluation Sizing of equipment in Figures 1 and 2 generally follows the guidelines in Turton et al. [15]. Distillation columns with sieve trays and degassers are considered as vertical pressure vessels. Column efficiency, height and diameter are calculated using the procedures described in Wankat [23]. Heat exchangers HE-104 in the DED-PSA process and HE-101 in the HPAD-PSA process are fixed tube shell-and-tube exchanger due to the need for large heat transfer area and lower cost. HE-109, HE-116 in the DED-PSA process and HE-106 in

10

the HPAD-PSA process require < 10 m2 of heat transfer area, and hence they are double pipe type. All remaining heat exchangers including condensers and reboilers are of floating-head shell and tube type. For sizing all exchangers in counter-current configuration, stream inlet/outlet temperatures and heat duty are taken from Aspen HYSYS simulation, and the required heat transfer area is calculated based on log mean temperature difference and overall heat transfer coefficient estimated by the resistances method described in Branan [24]. Appropriate values of film resistances and fouling factors are taken depending on stream conditions. All pumps are assumed to be centrifugal type except for P-103 (in Figure 1) and P-102 (in Figure 2), which are taken to be reciprocating type (because < 1 kW of shaft power is required). Compressors VC-101, VC-102 and VC-103 in both processes have suction pressures in the range of 25-82 kPa. Hence, they are assumed to be screw-type compressors [25], while the remaining compressors are of centrifugal type since their suction is at or above atmospheric pressure. All centrifugal compressors are assumed to have 75% adiabatic efficiency, whereas efficiency for VC-101, VC-102 and VC-103 are taken as 50%, 50% and 70% respectively, based on guidelines in Ryans and Bay [25]. For both DED-PSA and HPAD-PSA processes, material of construction is assumed to be stainless steel 304 for distillation columns, trays, degasser, compressors and pumps, and carbon steel for drives of compressors and pumps. In case of heat exchangers, material of construction for tubes is assumed to be stainless steel 304 except in HE-104, HE-105, and HE-116 of DED-PSA process and in HE-108 of HPAD-PSA process where it is assumed to be carbon steel. Material of construction of shells is assumed to be carbon steel except in HE-102 of DED-PSA and HPAD-PSA processes, where it is assumed to be stainless steel 304. These choices are made considering components, temperature and pressure of streams involved in each exchanger. FCI of both DED-PSA and HPAD-PSA processes is then calculated following the guidelines in Turton et al. [15]. Bare module cost of each equipment is updated to CEPCI of 600 as in [5]. Unit prices for electricity (16.8 $/GJ), low pressure steam (13.28 $/GJ), medium pressure steam (14.19 $/GJ) and cooling water (0.354 $/GJ) are taken from Turton et al. [15] and used for estimating AOC. Note that replacement cost of molecular sieves in PSA columns is not considered in AOC as their lifetime is 20 years as in [5]. Rather, it is included in FCI. Utilities used in the present study include low pressure steam (LPS) in the DC2 reboiler (HE-108) of HPAD-PSA 11

process, medium pressure steam (MPS) in HPC reboiler (HE-104) of DED-PSA process, electricity in the drives of compressors and pumps, and cooling water (wherever necessary in the heat exchangers of both processes). For HE-106 alone (in Figure 2), since the exit temperature (accounting for ∆Tmin of 10°C) of cooling water is less than 45°C (as specified in Turton et al. [15]), the unit price of cooling water is adjusted in proportion to the stated price of cooling water (entry and exit temperatures as 30 and 45°C respectively) in Turton et al. [15]. Finally, assuming payback period of FCI is 3 years, TAC is calculated by (1)

4. Preliminary design and economics of DED-PSA and HPAD-PSA processes Most parameters in the initial design of DED-PSA process (Figure 1) are taken from the optimal D-PSA configuration presented by Loy et al. [5]. For example, number of stages, feed stage and design specification values of LPC and DC2 are taken from D-PSA process. Besides, number of stages, feed stage and design specifications of HPC are chosen to minimize HPC reboiler duty and to keep condenser temperature of column HPC sufficiently high for heat integration with LPC reboiler. Desorption pressure and adsorption temperature of PSA are chosen in order to minimize the number of vacuum pumps (compressors) required to compress Stream 41 (in Figure 1). Finally, as depicted in Figure 4, feed split ratio to LPC and HPC is varied and chosen to minimize the hot utility required for the process. Table 2 summarizes parameters and their values used for the preliminary design of the DEDPSA process. Equipment sizing and economic evaluation of the process are carried out according to the procedures stated in Section 3. Table 3 shows the cost breakdown for DEDPSA process.

12

Figure 4: Hot utility requirement for the process versus feed split ratio (i.e., flow ratio between streams 2 and 1 in Figure 1). Filled marker corresponds to the optimal solution.

Table 2: Design parameters and their values for the DED-PSA process. Design variable

Value

Feed split ratio

0.41

Stream 6 temperature (°C)*

98.0

LPC pressure (kPa)

110

HPC pressure (kPa)

920

DC2 pressure (kPa)

102

Ethanol purity (wt.%) in Streams 12*, 25* and 56*

0.01

Ethanol purity (wt.%) in Stream 38*

92.6

Ethanol recovery (%) in Stream 15

99.4

Ethanol recovery (%) in Stream 34*

99.7

CO2 purity (ppm) in Stream 15

10

CO2 purity (ppm) in Stream 34*

23

PSA adsorption pressure (kPa)*

200

PSA desorption pressure (kPa)

25

Ethanol purity (wt.%) in stream 47*

99.8

*Values of these parameters are taken from Loy et al. [5].

13

Table 3: Cost breakdown for the DED-PSA process. Process equipment (cost in $ 1000) Pressure vessels (distillation columns, sieve trays and degasser)

5686.6

Compressor system (compressors and electric drives)

6338.0

Heat exchangers (including reboilers and condensers)

5388.8

PSA system (including two columns and adsorbent)

2629.9

Pumps

166.3

FCI ($ 1000)

20209.7

Process utilities (cost in $ 1000/year) Steam

7350.9

Electricity

799.2

Cooling Water

176.1

AOC ($ 1000/year)

8326.2

TAC ($ 1000/year)

15062.8

Unit cost of separation* ($/kg of bioethanol)

0.095

* Unit cost of separation is obtained by dividing TAC by bioethanol production of 159 million kg/year.

Similar to DED-PSA, most of the design parameters for HPAD-PSA process are taken from the optimal D-PSA process in Loy et al. [5]. The design specifications of DC2 (Figure 2) are changed since this column is simulated in the present study as a reboiled-absorber (equivalent to Reboiler Absorber module in Aspen HYSYS) plus an external condenser. On the other hand, in Loy et al. [5], DC2 is simulated as a distillation column with reboiler and condenser. For the base case design of HPAD-PSA process, ethanol concentration in DC2 bottoms is set to 0.008 wt.%, and the mass flow ratio between Streams 27 and 26 (hereafter referred as split ratio of Stream 26) is set to 0.71. Similar to the DED-PSA process, desorption pressure and adsorption temperature of PSA are chosen in order to minimize the number of vacuum pumps (compressors) required to compress the purge stream (i.e. Stream 35 in Figure 2). Table 4 summarizes the parameters and their values used for the preliminary design of the HPAD-PSA process. Economics of the HPAD-PSA preliminary design were evaluated for the two scenarios as explained previously in Section 2.2. Table 5 shows the cost breakdown for the HPAD-PSA process for the two scenarios. Scenario 1 is better than Scenario 2 and offers ~ 1% savings in TAC. Moreover, it has an additional degree of freedom (outlet pressure of K-101 in Figure 2), and so it is chosen for further analysis and comparison with DED-PSA process. Hereafter, Scenario 1 of HPAD-PSA process is simply referred to as HPAD-PSA.

14

Table 4: Design parameters and their values for the HPAD-PSA process Design variable

Value

Split ratio of Stream 26

0.71

DC2 pressure (kPa)

120

DC1 pressure (kPa)*

125

Ethanol purity (wt.%) in Stream 9*

0.01

Ethanol purity (wt.%) in Stream 22

0.008

Ethanol recovery (%) in Stream 17*

99.7

CO2 purity (ppm) in Stream 17*

23

Temperature of Stream 4 (°C)*

98.0

Adsorption pressure (kPa)*

200

Desorption pressure (kPa)

25

Ethanol purity (wt.%) in Stream 41*

99.8

*Values of these parameters are taken from Loy et al. [5].

Table 5: Cost breakdown for the HPAD-PSA process Process equipment (cost in $ 1000)

Scenario 1

Scenario 2

Pressure vessels (distillation columns, sieve trays and degasser)

5483.2

5483.2

Compressor system (compressors and electric drives)

13445.5

12894.9

Heat exchangers (including reboilers and condensers)

4470.8

4493.8

PSA system (including PSA columns and adsorbent)

3261.3

3256.0

Pumps FCI ($ 1000)

93.8 26754.6

93.8 26221.7

Process utilities ($ 1000/year) Steam

1185.9

1200.2

Electricity

2064.6

2334.0

Cooling Water

50.1

55.9

AOC ($ 1000/year) TAC ($ 1000/year)

3300.7 12218.9

3590.1 12330.7

Unit cost of separation ($/kg of bioethanol)

0.077

0.078

On comparing HPAD-PSA and DED-PSA processes, the former results in 60% savings in AOC and 32% additional capital investment over the latter; overall, the former reduces TAC by 19% (Tables 3 and 5). Savings in utilities, especially steam cost, in the HPAD-PSA process come from the vapor recompression of Stream 27 (Figure 2). This vapor 15

recompression increases condensation temperature of the stream, and thus enables its latent heat of condensation to be used in the reboilers (HE-103, HE-104 and HE-107) of DC1 and DC2. Therefore, hot utility (LPS) is required only for HE-108 (Figure 2). A priori, DED is expected to bring more savings. However, DED-PSA configuration is not attractive due to insufficient energy from the HPC condenser to LPC reboiler. There are two reasons for this: low distillate flowrates and vapor side draw in these columns. The vapor distillates withdrawn from LPC and HPC in the DED-PSA process consist mainly of CO2, which is present in the feed at very low concentrations (0.1 wt.%). Therefore, distillate rates are low when compared to bottoms or side draw, implying small condensation duty. Besides, using vapor side draw instead of liquid implies that the reboiler duty of the column will be higher than the condenser duty, as one expects larger vapor flowrates through the stages. In addition, due to high operating pressure of HPC, MPS is used in HE-104. All these result in 6 times steam cost for DED-PSA process compared to the cost of LPS for HPAD-PSA process; however, the latter requires 2.5 times electricity cost for compressors compared to the former (Tables 3 and 5). The increase in FCI from DED-PSA to HPAD-PSA may be counter intuitive at first. Indeed, DED-PSA configuration requires one extra distillation column and three additional heat exchangers, and thus one could expect FCI of this process to be higher. However, as can be seen in Tables 3 and 5, the cost of pressure vessels in the HPAD-PSA process is comparable to DED-PSA process. This is due to the higher number of stages and larger diameter of DC1 in the HPAD-PSA configuration. Also, capital investment for compressors in HPAD-PSA is twice as that for DED-PSA process. Compressor K-100 (in Figure 1) in the DED-PSA process compresses only the vapor distillate stream of DC, whereas this unit in the HPADPSA compresses all the vapor leaving the top stage of DC2 (in Figure 2). Besides, the latter process requires an extra compressor (K-101) for vapor recompression of Stream 27, leading to additional capital investment of ($ 4080103). Overall, savings in AOC (by 60%) outweigh additional capital investment (by 32%), and HPAD-PSA process offers 19% savings in TAC over DED-PSA process. Hence, MOO of HPAD-PSA process is investigated in the following sections. 5. Formulation and solution of the optimization problem MOO can provide quantitative trade-off between conflicting objective functions, which underpins the choice of an optimal design for the proposed process. For optimizing the 16

HPAD-PSA process, we define the objective functions, identify decision variables and constraints, and outline the optimizer used, in this section. Objectives: The objective functions chosen in the present study are minimization of both FCI and AOC. In the design of chemical plants, FCI and AOC are usually two conflicting objectives, as higher investments for the process equipment will usually result in more savings in AOC. Therefore, these two objectives are expected to generate a set of nondominated optimal solutions. Decision variables: Table 6 lists the key decision variables influencing both the objectives. Bounds on these decision variables are chosen to guarantee convergence and consequently the robustness of process simulation in Aspen Plus, throughout MOO. Further, values outside the specified bounds may affect the HEN topology thus resulting in extra heat exchangers required or approach temperature falling below 10°C for some heat exchangers. An initial trial of MOO was performed with maximum number of generations (MNG) as 50 and a smaller population size of 20, in order to test the bounds on decision variables chosen. It was found that some of the chosen decision variables, namely, split ratio of stream 18 and K-100 outlet pressure resulted in values corresponding to their upper bounds of 0.714 (0.671-.0714) and 360 kPa (221-360) respectively. In addition, K-101 outlet pressure (474660 kPa) was close to its lower bound (474 kPa). Based on these observations, the values of split ratio of stream 18, K-100 outlet pressure and K-101 outlet pressure were set to 0.714, 360 kPa and 474.4 kPa respectively. Thus, the benefit of performing an initial trial MOO with a smaller population size and MNG helps to reduce number of decision variables and/or narrow the bounds on each decision variable, thus reducing the computational time of subsequent MOO. Constraints: Ethanol purity in the inlet stream to the PSA system should be greater than 80 wt.% and minimum temperature approach in the HEN higher than 10°C. The first constraint comes from the study of Loy et al. [5], whereby it is treated as a decision variable, i.e., ethanol purity in DC2 distillate stream and its lower bound is 80 wt.%. In this work, this ethanol purity changes according to split ratio of Stream 26 and a constraint on it is included in order to ensure a feasible adsorption (Table 6). All governing equations of each unit operation in the process are satisfied through the simulation in Aspen HYSYS. Hence, they are not included as equality constraints in MOO problem.

17

Table 6: MOO problem formulation for HPAD-PSA process Objective Functions Objective function 1 ($ 1000)

Minimize FCI

Objective function 2 ($ 1000/year)

Minimize AOC

Decision Variables

LB

UB

Temperature of stream 4 (°C)

98.0

98.5

Number of ideal stages in DC1

30

35

Number of ideal stages in DC2

16

24

DC1 recycle feed stage*

0.00

0.10

DC2 recycle feed stage*

0.26

0.82

CO2 concentration in stream 17 (ppm)

10

31

Ethanol concentration in stream 9 (ppm)

70

180

Ethanol concentration in stream 22 (ppm)

50

100

Constraints Ethanol purity in PSA feed (wt.%)

≥ 80

Minimum approach temperature, Tmin in the HEN (°C)

≥ 10

* Feed stage is given as a fraction of the total number of stages,

, where i is the actual feed

stage (in Aspen HYSYS) and N is the number of stages (in Aspen HYSYS). For process simulation in Aspen HYSYS, this fraction is multiplied by (N -1), 1.0 is added to the result and then rounded off to an integer value for feed stage.

Optimizer: For MOO, the Integrated Multi-Objective Differential Evolution (IMODE) program from [14] is used. This algorithm is based on multi-objective differential evolution incorporating tabu list and self-adaptation of algorithm parameters. It also includes two progress-based termination criteria besides MNG. MOO was performed for a population size of 100 and MNG of 100 as the termination criterion. Initial values for crossover and mutation probabilities were both fixed as 0.5, and values of other IMODE parameters were taken from [14]. 6. Results and discussion 6.1 MOO results for the HPAD-PSA process Figure 5 shows the Pareto-optimal front obtained from MOO with a population of 100 and MNG of 100. In order to confirm the convergence to the Pareto-optimal front, intermediate results are plotted in the top right corner at an interval of 20 generations. As it can be observed, there is very little improvement in the values of the objective functions in their respective Pareto fronts after 60 generations, thus indicating that the optimization has 18

converged at the end of 100 generations. However, with increasing number of generations, more number of optimal solutions may be obtained. Based on this observation, one can choose to terminate the program any time after 60 generations to reduce the computational time. In the Pareto-optimal front in Figure 5, FCI ranges from 23.99 to 24.35 million $ and AOC from 3.20 to 3.25 million $/year. The optimal solution having the lowest TAC (11.24 million $/year), shown as an open circle in Figure 5, is selected for further discussion and comparison with other studies. Compared to the preliminary design of HPAD-PSA, this optimal solution brings 8% savings in TAC. The narrow range of FCI (~400,000) and AOC (~ 50,000) in Figure 5 is possibly due to the non-conflicting effect of many decision variables on the objectives. Indeed, for some decision variables (not shown in Figure 6 for brevity), namely, number of ideal stages in DC1, DC1 recycle feed stage, DC2 recycle feed stage, and ethanol concentration in stream 22 were approximately the same for all optimal solutions in Figure 5. This means that each of these decision variables has a value minimizing both FCI and AOC. Further, optimal values of CO2 concentration in stream 17 do not present a clear trend, indicating that this decision variable doesn’t have much effect on the objectives. Optimal values of Ethanol concentration in stream 9 also present a scattered pattern, varying between 120 and 180 ppm. Variation of the objective functions with respect to the remaining two decision variables is shown in Figure 6. The trends observed in Figures 6a-b can be understood from the HEN presented in Appendix A. Indeed, by further preheating the feed stream to DC1 to increase Stream 4 temperature, the (total) duty of DC1 reboiler will decrease. As a result, HE-103 duty (Figure A1) will be smaller, and consequently more heat will be available in stream 20 to boil up the bottoms of DC2 (HE-107). Therefore, the duty of HE-108 decreases, and consequently required LPS and its cost decrease. Increase in FCI results mainly from the increase in the area of HE-102. The sharp change in FCI and AOC at Stream 4 temperature of 98.5 oC is due to a corresponding increase in DC2 no. of stages. Trends shown in Figures 6c-d can be explained as follows. Increase in number of stages results in higher cost of DC2 and lower steam consumption in DC2 reboiler (HE-108).

19

Figure 5: Pareto-optimal front for the HPAD-PSA process for a population of 100 and MNG of 100. Unfilled marker in the main plot corresponds to the optimal solution with the lowest TAC. Top right plot shows MOO progression at different generations.

Figure 6: Variation of FCI and operating cost with certain decision variables, corresponding to the Pareto-optimal front in Figure 5.

20

Optimal values of design parameters, process economics and key performance indicators, namely, specific energy consumption (SEC) and unit cost of separation for the chosen optimal solution for HPAD-PSA process are summarized in Table 7. SEC (MJ-fuel/kg bioethanol) is calculated according to the following equation:

(2)

where bioethanol production rate is 159 million kg/year or 5.353 kg/s (assuming plant operation of 8250 hours per year), and 0.9 and 0.3 are efficiency of steam and electricity production, respectively. As shown in Table 7, capital cost of compressors accounts for 54% of FCI. The electricity consumed in the compressors contributes to > 99% of total electricity cost and 65% of AOC. This shows that the compressors are the most expensive units in this process. However, they are also responsible for large savings in total utility cost, as vapor recompression brings significant reduction in steam consumption. The unit cost of electricity given in Section 3 (16.8 $/GJ) may vary depending on the power generation system supplying electricity to the plant. For example, estimated unit price of electricity for a nuclear power plant is between 0.112 $/kWh and 0.183 $/kWh [26]. If the HPAD-PSA process is for a plant getting all electricity from a nuclear power plant at unit cost of 0.15 $/kWh (i.e., 41.7 $/GJ), electricity cost of the HPAD-PSA process increases to 5,235,600 $/y. Ultimately, this increases unit TAC to 0.090 $/kg of bioethanol from 0.071 $/kg of bioethanol given earlier in Table 7.

21

Table 7: Design parameters, process economics and key performance indicators for the chosen optimal solution for the HPAD-PSA process. Design variables Column diameter (m) Column height (m) Number of ideal stages Feed stage Side-draw stage Recycle stage Column pressure (kPa) Condenser duty (MW) Reboiler duty (MW) Compressor shaft power (MW)c Degasser column diameter (m) Degasser column height (m) PSA column diameter (m) PSA column height (m) Amount of adsorbent used in two PSA columns (kg) Process economics (in $ 1000 or $ 1000/year) Pressure vessels (distillation columns, sieve trays and degasser) Compressor system (compressors and electric drives) Heat exchangers (including reboilers and condensers) PSA system (including PSA columns and adsorbent) Pumps Total FCI LP steam DC1 DC2 Electricity K-101 K-102 K-103 VC-101 VC-102 VC-103 Pumps Cooling water Total AOC Key performance indicators TAC (in $ 1000/year) Thermal energy usage (MW) Electrical energy usage (MW) Specific energy consumption (MJ-fuel/kg bioethanol) Unit cost of separation ($/kg of bioethanol)d

DC1 2.4 33.5 30 4 7 2 125 0.85 18.7a

DC2 3.2 21.3 18 14 12 120 18.4 2.30a + 2.71b 4.0 0.8 2.3 1.8 15.1 37449 4846.7 12913.3 4503.5 1739.3 93.8 24096.6 1068.0 1068.0

2094.3 1340.2 267.6 96.2 170.9 116.5 99.6 3.4 49.3 3211.6 11243.8 2.71 4.23 3.19 0.071

a) Energy supplied from hot streams within the process. b) External energy supplied (steam) after accounting for heat integration within the process. c) Directly obtained from Aspen HYSYS; needs to be divided by motor efficiency of 0.95 to find the electrical energy required. d) Unit cost of separation =

22

6.2 Comparison with other processes In order to compare the proposed HPAD-PSA process with other bioethanol separation methods, SEC is chosen as the criterion. Owing to differences (e.g., cost correlations, equipment type, data used etc.), comparison in terms of TAC or unit production cost in reported studies may not be on the same basis, and so cost comparison is not performed. Since the present work proposes alternative technologies in the pre-concentration section of bioethanol separation studied by Loy et al. [5], the first step is to compare the bioethanol separation methods evaluated in this reference. Loy et al. [5,27] investigated three technologies namely, D-PSA (described earlier in Section 2.1), E-DWC technology (singlecolumn configuration of Kiss and Ignat [4]) using ethylene glycol (EG) as solvent, and conventional three column extractive distillation (3-column ED) using EG, all three for bioethanol separation from a ternary feed of ethanol, water and CO 2. Note that E-DWC has 98% recovery of EG compared to 99.99% recovery in 3-column ED. Refer to Loy et al. [5,27] for elaborate discussion/comparison among D-PSA, E-DWC and 3-column ED. As evident from Figure 7, HPAD-PSA process requires 38%, 41% and 55% lesser energy than D-PSA, E-DWC and 3-column ED respectively, primarily due to the decrease in hot utility demand. Note that the reboiler duty of DC1 reported in Table 7 of Loy et al. [5] for D-PSA process includes thermal energy for feed preheating, which should not be added to reboiler duty because it is supplied from another process stream. So, upon deducting the feed preheating duty (of 32.2 GJ/h) from the reported value, DC1 reboiler duty becomes 73.8 GJ/h. For DC2, the reported reboiler duty is supplied by thermal energy from vapor recompression technology. Similarly, the reboiler duty of pre-concentration column (PDC) reported in Table 1 of Loy et al. [27] for 3-column ED process includes thermal energy for feed preheating, which should not be added for the same reason. So, upon deducting the feed preheating duty (of 46.2 GJ/h) from the reported value, PDC reboiler duty becomes 64.8 GJ/h. Further, for evaluating SEC, thermal energy in eq. 2 only accounts thermal energy supplied using a hot utility (e.g. steam) and not from heat integration. Hence, values of thermal energy demand for D-PSA and 3-column ED used in evaluating SEC are 73.8 and 115.3 GJ/h respectively; the second value includes reboiler duty of PDC, extractive distillation and solvent recovery columns in 3-column ED. Energy savings (as a result of lower SEC) with E-DWC over 3-column ED process in Figure 7 is due to relatively lower (98%) solvent recovery in the former process. Upon increasing EG recovery to 98.5%,

23

simulation results show that SEC of E-DWC increases to 6.5 MJ-fuel/kg, closer to the value of 7.1 MJ-fuel/kg for 3-column ED.

Figure 7: Comparison of SEC for HPAD-PSA process with bioethanol separation processes studied by Loy et al. [5,27]. Further comparison with other recent studies is made based on the papers reviewed in Singh and Rangaiah [1] for bioethanol recovery and dehydration. For a consistent comparison, these studies are selected using the following criteria: processes using feed with ethanol concentration in the range 10-12 wt.% and targeting ethanol purity ≥ 99.6 wt.% in the product. The selected technologies are briefly outlined here. Luyben [28] simulated heterogeneous azeotropic dehydration (AD) of ethanol in a 3-column sequence (hereby referred to as 3-column AD), using both benzene and cyclohexane as the entrainer. A tradeoff between the distillate composition from PDC and the overall energy required in the process is established, with an optimal ethanol concentration as 80 mol% (91 wt. % ethanol), considerably far from azeotropic concentration of ~ 95 wt. % ethanol. Kiss and Ignat [4] proposed ED in a dividing-wall column (E-DWC) and optimized it for thermal energy required. Errico et al. [29] studied many alternative sequences of 4-column ED (i.e., additional PDC followed by 3-column ED) by introducing various combinations of partial condensers and merging of distillation columns utilized to perform the same separation task. Among all the alternative sequences studied, 3-column ED (having 2 partial condensers) with vapor recycle (i.e. recycling of vapor distillate from SRC to PDC) achieved 22% energy savings over 4-column ED. Later, Errico et al. [30] extended the same principle

24

to obtain intensified 2-column sequence (PDC followed by EDWC with vapor recycle); this sequence resulted in marginal energy savings of < 2% over 3-column ED with vapor recycle. Kiss and Ignat [31] studied the conventional 3-column ED sequence, optimizing the process with respect to the distillate composition from the pre-concentration column. Sensitivity analysis indicates that a distillate composition of 91 wt. % ethanol leads to optimum TAC along with lower energy requirement, similar to the findings for the 3-column AD [28]. Gudena et al. [32] studied the MOO of a stripper-vapor permeation (S-VP) process (same as the configuration reported in [33]) for different sets of objectives, namely, bioethanol purity and operating cost, and ethanol loss and operating cost. Vazquez-Ojeda et al. [34] suggested the use of liquid-liquid extraction prior to the pre-concentration column in the conventional 3-column ED sequence. However, for a feed range of 5-12 wt. % ethanol, they found that the conventional 3-column ED is better than the alternative configuration in terms of operating cost. Later, Errico et al. [35] studied steady-state and dynamic behavior of 2 and 3-column ED sequences. On comparing, 2-column ED sequence with water side-draw from the first column reduces energy consumption and capital cost while improving control properties. Luo et al. [3] investigated HP assisted E-DWC process to improve the energy efficiency of E-DWC process studied by Kiss and Ignat [4]. The compressed vapor stream is used to provide thermal energy to a side-reboiler, resulting in about 40% savings in energy. Recently, Singh and Rangaiah [8] studied MOO of D-VP process using a ternary feed of ethanol, water and CO2 to produce dehydrated bioethanol using unit COM and greenhouse gas emissions as objective functions. Figure 8 compares the above described studies in terms of SEC; the studies are categorised based on ethanol concentration in the feed and the product, considering technologies with solvent loss < 1.5%. Note that under any category, based on our simulations for D-PSA, HPAD-PSA, 3-column ED and E-DWC processes, SEC is found to vary marginally for a change in ethanol feed and product concentrations of < 2 wt. % and < 0.5 wt. %, respectively. Thus, this forms the basis for comparing different processes. As expected and as can be seen in Figure 8, conventional 3-column ED and AD processes without heat integration require more energy (7.9-8.8 MJ-fuel/kg bioethanol) than other processes. Modifications to conventional 3-column ED in the form of vapor recycling from SRC to PDC, PDC followed by EDWC with vapor recycle or incorporating HP to E-DWC enhances the energy efficiency with SEC ranging 4.9-5.0 MJ-fuel/kg bioethanol. 25

Figure 8: Comparison amongst different bioethanol separation processes in the recent literature, using SEC as the criterion. Hybrid separation technologies involving membranes are also found to improve the energy efficiency of bioethanol separation with SEC ranging 5.4-6.3 MJ-fuel/kg bioethanol. The proposed HPAD-PSA process is the most energy-efficient process amongst the separation processes compared in Figures 7 and 8. Its SEC of 3.2 MJ-fuel/kg of bioethanol corresponds to reduction of 35-64% compared to other, recent bioethanol separation processes. 7. Conclusions This study explores bioethanol recovery and dehydration using DED-PSA and HPAD-PSA processes, for the first time. For bioethanol recovery, DED and HPAD technologies are proposed as alternatives to simple distillation in order to improve the energy efficiency, prior to bioethanol dehydration by PSA, the widely used technology in the industry. In the preliminary design and analysis, HPAD-PSA process resulted in 19% savings in TAC compared to DED-PSA, primarily due to 60% savings in AOC, despite 32% additional 26

capital investment. Hence, MOO of HPAD-PSA is performed by simultaneously minimizing FCI and AOC as the objective functions. An optimal solution having the lowest TAC is chosen from the resulting Pareto-optimal front, and it leads to additional 8% savings in TAC compared to the preliminary design. Lastly, the proposed HPAD-PSA process is compared with other bioethanol separation technologies; it shows that HPAD-PSA process is the most energy-efficient, and also results in 38% savings in SEC as compared to D-PSA process used in US industries. Hence, HPAD-PSA is very promising for both new bioethanol plants and retrofitting existing plants. Therefore, control of this process should be studied for its acceptance and implementation in the industry. Acknowledgements The first and the second authors are grateful for the financial support provided by the National University of Singapore under Graduate Student Researcher scheme.

References [1]

A. Singh, G.P. Rangaiah, Review of Technological Advances in Bioethanol Recovery and Dehydration, Ind. Eng. Chem. Res. 56 (2017) 5147–5163.

[2]

M.A.M. Hussain, P.H. Pfromm, Reducing the Energy Demand of Cellulosic Ethanol through Salt Extractive Distillation Enabled by Electrodialysis, Sep. Sci. Technol. 48 (2013) 1518–1528.

[3]

H. Luo, C.S. Bildea, A.A. Kiss, Novel heat-pump-assisted extractive distillation for bioethanol purification, Ind. Eng. Chem. Res. 54 (2015) 2208–2213.

[4]

A.A. Kiss, R.M. Ignat, Innovative single step bioethanol dehydration in an extractive dividing-wall

column,

Sep.

Purif.

Technol.

98

(2012)

290–297.

http://dx.doi.org/10.1016/j.seppur.2012.06.029. [5]

Y.Y. Loy, X.L. Lee, G.P. Rangaiah, Bioethanol recovery and purification using extractive dividing-wall column and pressure swing adsorption: An economic comparison after heat integration and optimization, Sep. Purif. Technol. 149 (2015) 413–427. http://dx.doi.org/10.1016/j.seppur.2015.06.007.

[6]

L.C. Nhien, N.V.D. Long, M. Lee, Novel heat–integrated and intensified biorefinery process for cellulosic ethanol production from lignocellulosic biomass, Energy 27

Convers.

Manag.

141

(2017)

367–377.

http://dx.doi.org/10.1016/j.enconman.2016.09.077. [7]

C.E. Torres-Ortega, B.G. Rong, Synthesis and Simulation of Efficient Divided Wall Column Sequences for Bioethanol Recovery and Purification from an Actual Lignocellulosic Fermentation Broth, Ind. Eng. Chem. Res. 55 (2016) 7411–7430.

[8]

A. Singh, G.P. Rangaiah, Process Development and Optimization of Bioethanol Recovery and Dehydration by Distillation and Vapor Permeation for Multiple Objectives, in: G.P. Rangaiah (Ed.), Differ. Evol. Chem. Eng. - Dev. Appl., World Scientific, Singapore, 2017. https://doi.org/10.1142/9789813207523_0010.

[9]

Ö. Yildirim, A.A. Kiss, E.Y. Kenig, Dividing wall columns in chemical process industry: A review on current activities, Sep. Purif. Technol. 80 (2011) 403–417.

[10] A.A. Kiss, Advanced Distillation Technologies - Design, Control and Applications, Wiley, Chichester, 2013. [11] R. Wooley, M. Ruth, J. Sheehan, K. Ibsen, H. Majdeski, A. Galvez, Lignocellulosic Biomass to Ethanol Process Design and Economics Utilizing Co-Current Dilute Acid Prehydrolysis and Enzymatic Hydrolysis Current and Futuristic Scenarios, Golden, 1999. http://www.osti.gov/servlets/purl/12150-rKCQEi/native/. [12] D. Humbird, R.E. Davis, L. Tao, C.M. Kinchin, D.D. Hsu, A. Aden, P. Schoen, J. Lukas, B. Olthof, M. Worley, D. Sexton, D. Dudgeon, Process Design and Economics for Biochemical Conversion of Lignocellulosic Biomass to Ethanol, Golden, 2011. http://www.nrel.gov/biomass/pdfs/47764.pdf. [13] F. Tavares, J. Pascoal, B. Maia, Expanding Ethanol Dehydration, (2018). http://www.ethanolproducer.com/articles/6434/expanding-ethanoldehydration (accessed April 18, 2018). [14] S. Sharma, G.P. Rangaiah, F. Marechal, Integrated Multi-Objective Differential Evolution and its Application to Amine Absorption Process for Natural Gas Sweetening, in: G.P. Rangaiah, S. Sharma (Eds.), Differ. Evol. Chem. Eng., World Scientific, Singapore, 2017. [15] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, D. Bhattacharyya, Analysis, Synthesis, and Design of Chemical Processes, 4th ed., Pearson, Upper Sadle River, 2013.

28

[16] W.D. Seider, J.D. Seader, D.R. Lewin, S. Widagdo, Product and Process Design Principles - Synthesis, Analysis and Evaluation, 3rd ed., Wiley, Hoboken, 2010. [17] I.C. Kemp, Pinch analysis and process integration: A user guide on process integration for the efficient use of energy, 2nd ed., Butterworth-Heinemann, Oxford, 2007. [18] C.L. Smith, Distillation Control: An Engineering Perspective, Wiley, Hoboken, 2012. [19] I. Patraşcu, C.S. Bildea, A.A. Kiss, Dynamics and control of a heat pump assisted extractive dividing-wall column for bioethanol dehydration, Chem. Eng. Res. Des. 119 (2017) 66–74. doi:10.1016/j.cherd.2016.12.021. [20] W.L. Luyben, Series versus parallel reboilers in distillation columns, Chem. Eng. Res. Des. 133 (2018) 294–302. doi:10.1016/j.cherd.2018.03.025. [21] I. Dalmolin, E. Skovroinski, A. Biasi, M.L. Corazza, C. Dariva, J.V. Oliveira, Solubility of carbon dioxide in binary and ternary mixtures with ethanol and water, Fluid Phase Equilib. 245 (2006) 193–200. [22] M.D. Koretsky, Engineering and Chemical Thermodynamics, 2nd ed., Wiley, Hoboken, 2013. [23] P.C. Wankat, Separation Process Engineering, 2nd ed., Pearson, Westford, 2007. [24] C. Branan, Rules of Thumb for Chemical Engineers, 4th ed., GPP, Burlington, 2005. [25] J. Ryans, J. Bays, Run clean with dry vacuum pumps, Chem. Eng. Prog. 97 (2001) 32–41. [26] Lazard,

Levelized

Cost

of

Energy

Analysis

-

Version

11.0,

2017.

https://www.lazard.com/media/450337/lazard-levelized-cost-of-energy-version110.pdf. [27] Y.Y. Loy, X.L. Lee, G.P. Rangaiah, Optimization and Economic Evaluation of Bioethanol Recovery and Purification Processes involving Extractive Distillation and Pressure Swing Adsorption, in: K. V. Gernaey, J.K. Huusom, R. Gani (Eds.), 25th Eur. Symp. Comput. Aided Process Eng., Elsevier Inc., Copenhagen, 2015: pp. 413–418. [28] W.L. Luyben, Economic optimum design of the heterogeneous azeotropic dehydration of ethanol, Ind. Eng. Chem. Res. 51 (2012) 16427–16432. [29] M. Errico, B.G. Rong, G. Tola, M. Spano, Optimal Synthesis of Distillation Systems 29

for Bioethanol Separation. Part 1: Extractive Distillation with Simple Columns, Ind. Eng. Chem. Res. 52 (2013) 1612–1619. [30] M. Errico, B.G. Rong, G. Tola, M. Spano, Optimal Synthesis of Distillation Systems for Bioethanol Separation. Part 2. Extractive Distillation with Complex Columns, Ind. Eng. Chem. Res. 52 (2013) 1620–1626. [31] A.A. Kiss, R.M. Ignat, Optimal Economic Design of an Extractive Distillation Process for Bioethanol Dehydration, Energy Technol. 1 (2013) 166–170. [32] K. Gudena, G.P. Rangaiah, S. Lakshminarayanan, Multi-Objective Optimization of a Hybrid Steam Stripper-Membrane Process for Continous Bioethanol Purification, in: G.P. Rangaiah, A. Bonilla-Petriciolet (Eds.), Multi-Objective Optim. Chem. Eng., Wiley, Chichester, 2013. [33] Y. Huang, R.W. Baker, L.M. Vane, Low-Energy Distillation-Membrane Separation Process, Ind. Eng. Chem. Res. 49 (2010) 3760–3768. [34] M. Vázquez-Ojeda, J.G. Segovia-Hernández, S. Hernández, A. Hernández-Aguirre, A.A. Kiss, Design and optimization of an ethanol dehydration process using stochastic methods, Sep. Purif. Technol. 105 (2013) 90–97. [35] M. Errico, C. Ramírez-Márquez, C.E. Torres Ortega, B.G. Rong, J.G. SegoviaHernandez, Design and control of an alternative distillation sequence for bioethanol purification, J. Chem. Technol. Biotechnol. 90 (2015) 2180–2185.

30

Appendix A Heat integration network (HEN) is partially omitted in Figures 1 and 2 for clarity of process flowsheet. Figures A1 and A2 present the HEN for the DED-PSA and HPAD-PSA processes, respectively. Tables A1 and A2 give duties of heat exchangers in Figures A1 and A2, respectively. In these tables, Name refers to the heat exchanger in Figure 1 or 2.

Figure A1: HEN for the DED-PSA process in Figure 1. Numbers on the left of the diagram refers to stream numbers as indicated in Figure 1. Numbers above the line are stream temperatures in oC.

31

Table A1: HEN data for the DED-PSA process in Figure 1. No.

Name

1 2 3 4 5 6 7 8

HE-105 HE-103 HE-106 HE-107 HE-108 HE-109 HE-110 HE-111

Duty (kW) 7,178 9,199 297 532 982 17 14,127 206

Heating/cooling No. medium Process stream 9 Process stream 10 Process stream 11 Process stream 12 Cooling water 13 Cooling water 14 Cooling water 15 Cooling water 16

Name HE-112 HE-113 HE-114 HE-102 HE-115 HE-101 HE-104 HE-116

Duty (kW) 303 258 4,769 419 873 6,563 17,384 92

Cooling/heating medium Cooling water Cooling water Process stream Process stream Cooling water Process stream MPS LPS

Figure A2: HEN for the optimal HPAD-PSA process in Figure 2. Numbers on the left of the diagram refers to stream numbers as indicated in Figure 2. Numbers above the line are stream temperatures in oC.

32

Table A2: HEN data for the HPAD-PSA process in Figure 2. No. 1 2 3 4 5 6 7 8 9 10 11 12 13

Name HE-105 HE-106 HE-101 HE-103 HE-107 HE-102 HE-109 HE-110 HE-111 HE-112 HE-104 HE-113 HE-108

Duty (kW) 846 1 15,291 13,873 2,301 626 1,625 219 329 287 4,876 1,379 2,708

33

Cooling/Heating medium Cooling water Cooling water Process stream Process stream Process stream Process stream Cooling water Cooling water Cooling water Cooling water Process stream Cooling water LPS

CO2 purge

HE-103b

HE-102b HE-106 HE-107b

Degasser

K-101

HE-109 HE-105

K-100

HE-104b

HE-113

K-102 HE-110

Product Ethanol: 99.8 wt.%

HE-101

Water:    0.2 wt.% HE-102a

DC2

DC1

PSA

Feed Ethanol: 10.0 wt.% Water:    89.9 wt.% CO2:       0.1 wt.%

HE-103a

Specific energy consumption: 3.2 MJ-fuel/kg of bioethanol Unit cost of production: 0.071 USD/kg of bioethanol

HE-104a HE-107a

HE-108

Production rate: 159,000 t/y Ethanol recovery: 99.7%

Waste water VC-103

HE-112

VC-102

HE-111

VC-101

Heat-Pump Assisted Distillation versus Double-Effect Distillation for Bioethanol Recovery followed by Pressure Swing Adsorption for Bioethanol Dehydration

Ashish Singh, Sergio da Cunha, G.P. Rangaiah*

HIGHLIGHTS 

Studied distillation-pressure swing adsorption (D-PSA) for bioethanol separation



Analyzed two alternatives for improving energy efficiency of D-PSA process



Heat-pump assisted distillation (HPAD) and double-effect distillation (DED)



HPAD-PSA outperforms DED-PSA in terms of total annual cost (~19% savings)



HPAD-PSA is the most efficient with 38% lower energy than D-PSA used in industries

34