1 Feasibility Assessment of the Production of Bioethanol from Lignocellulosic Biomass Pretreated with Acid Mine Drainage (AMD) Nicholas W. Burman, Craig Sheridan, Kevin G. Harding* Industrial and Mining Water Research Unit; Centre in Water Research and Development; School of Chemical and Metallurgical Engineering All of the University of the Witwatersrand, Johannesburg, Private Bag 3, Wits, 2050, South Africa *Kevin.Harding@wits.ac.za Declaration of Interest: none A techno-economic evaluation of a lignocellulosic bioethanol facility that uses acid mine drainage for the pre-treatment of weeping love grass (Eragrostis curvula) was performed. Both separate hydrolysis and fermentation (SHF) and simultaneous saccharification and fermentation (SSF) reactor configurations were evaluated. Results were compared to an evaluation of the same process with biomass pre-treated with dilute H2SO4. Capital and operating costs were estimated and a simple economic evaluation was conducted. It was found that all scenarios made a loss except for biomass pre-treated with H2SO4 in the SHF reactor configuration, although the high capital cost resulted in a payback period of 80.7 years, which is unfeasible. SHF was found to produce more ethanol at a lower capital cost than SSF, indicating that it is more economically feasible. Incorporating the remediation of AMD into a simultaneous process could help improve process economics. It is thus recommended that a techno-economic evaluation be conducted on a process that produces bioethanol through SHF and simultaneously remediates AMD. Keywords: Lignocellulosic Bioethanol; Economic Evaluation; Acid Mine Drainage (AMD); Pre- treatment; Separate Hydrolysis and Fermentation (SHF); Simultaneous Saccharification and Fermentation (SSF) 1 Introduction The Intergovernmental Panel on Climate Change has concluded in its 5th Assessment Report that the burning of fossil fuels is “extremely likely” to be the cause of recent increase in atmospheric and oceanic temperatures [1]. Although starch-based bioethanol has been demonstrated to be a renewable alternative to petrol [2], it has various drawbacks including 2 (amongst others) an increase in food prices, net energy losses, a reduction of available land for food production and land degradation [3]. Lignocellulosic biomass (agricultural waste, forestry residue and energy crops) has been identified as a source of biomass that can potentially be used to produce bioethanol without these drawbacks in an environmentally friendly manner [4]. Lignocellulosic biomass consists of three fractions, cellulose fibrils (linearly linked glucose molecules) surrounded by a matrix of hemicellulose and lignin [5]. In the production of bioethanol, cellulose is hydrolyzed to glucose by cellulase enzymes. Glucose is then fermented to produce bioethanol [6]. Unfortunately, the presence of the lignocellulosic matrix prevents cellulase enzymes from accessing cellulose fibrils, hindering enzymatic hydrolysis. To overcome this the ligno-10 hemicellulosic matrix needs to be disrupted in a pre-treatment process. Various pre-11 treatment processes have been investigated including physical, chemical, biological and 12 physico-chemical. Dilute H2SO4 pre-treatment has been found to be the most commercially 13 favorable process [5,7–10]. 14 Another drawback to enzymatic hydrolysis is that the glucose produced inhibits cellulase 15 activity. This is most evident in the separate hydrolysis and fermentation (SHF) reactor 16 configuration, in which enzymatic hydrolysis and fermentation occur in separate reactors. 17 To overcome this various reactor configurations, which aim to maintain a low glucose 18 concentration, have been investigated. These include; simultaneous saccharification and 19 fermentation (SSF), in which enzymatic hydrolysis and fermentation occur simultaneously in 20 the same reactor; simultaneous saccharification and co-fermentation (SSCF) in which 21 enzymatic hydrolysis of cellulose, fermentation of glucose and co-fermentation of 22 hemicellulose produced in pre-treatment all occur in the same reactor; and consolidated 23 3 bioprocessing (CBP) in which enzyme production, enzymatic hydrolysis, and fermentation all 1 occur in the same reactor, with a host of different micro-organisms. 2 Recently there have been investigations into the use of acid mine drainage (AMD) for the 3 pre-treatment of lignocellulosic biomass [11,12]. The H+ ions present in the AMD catalyzes 4 the hydrolyzes of hemicellulose, releasing sugars into the AMD, similar to dilute H2SO4 pre-5 treatment. 6 AMD is formed through a series of geochemical reactions when sulfate-rich minerals are 7 exposed to oxygen and water, often through mining activity [13]. It is highly acidic (pH 1-3) 8 with high sulfate (<20 g/L) and heavy metal (<5 g/L) concentration [14]. AMD has been 9 found to harm aquatic and riparian ecosystems due to the toxic effects of increased 10 concentration of acid, sulfate and heavy metals, on aquatic life. As the pH rises through 11 natural processes dissolved heavy metals precipitate on the ecosystem floor [15]. 12 Recently, there has been a study that determined the optimal pre-treatment times for 13 biomass pre-treated with AMD, as well as the rate of enzymatic hydrolysis, fermentation, 14 and SSF of this optimally pre-treated biomass [16]. 15 It has also been shown that the sugars present on AMD after pre-treatment can be used by 16 sulfate-reducing bacteria (SRB) in a process known as dissimilatory sulfate reduction [11,17]. 17 This process partially remediates the water through a reduction of sulfate and heavy metal 18 concentrations and an increase in pH. Although various work has been done on the 19 simultaneous production of ethanol and remediation of AMD [10–12,17–19], there has yet 20 to be a techno-economic evaluation of any aspect of this process. 21 There have been various techno-economic evaluations performed on the production of 22 bioethanol from lignocellulosic biomass [20]. Although all of these studies were based on 23 4 enzymatic hydrolysis and fermentation of lignocellulosic biomass, different processes and 1 feedstocks have been investigated. Feedstocks evaluated included corn stover [21–25], 2 switchgrass [20,26,27], hardwoods and softwoods [20,26,28,29] and paper sludge [30]. 3 There have been various pre-treatments evaluated including, dilute acid pre-treatment 4 [20,21,25,26,28], ammonia fiber explosion (AFEX) [23,24,27] and hot water [29]. Different 5 reactor configurations for the enzymatic hydrolysis and fermentation process have been 6 evaluated, including SHF [21], SSF [30], CBP [23,27] and SSCF [23]. 7 Most of these techno-economic studies have been performed using a variety of process 8 simulation, flow sheeting and engineering software, e.g. Aspen Plus®, MS Excel, and 9 Capcost. Typically, flowsheets and stream tables are first developed, after which they are 10 used to provide estimates of operating and capital costs which are used in a techno-11 economic evaluation of the process. 12 The study presented here investigated the production of bioethanol from Eragrostis curvula 13 (Weeping lovegrass), pre-treated with AMD, and for comparison, pre-treated with H2SO4. 14 Both SHF and SSF reactor configurations were investigated. The AMD remediation process is 15 not investigated in this study. 16 The objectives of this study are to: develop process flow sheets for all scenarios; estimate 17 the operating costs of the process; estimate the capital costs of the process, and; determine 18 the economic feasibility through economic evaluation. 19 2 Methods 20 2.1 Process flowsheet development 21 2.1.1 Simulation Basis 22 5 The simulation was carried out using Aspen Plus® V8.4 (www.aspentech.com), with a basis 1 of 90 ton/hr of milled (4mm) dry Eragrostis curvula (weeping lovegrass) grass with a 2 composition of 37% cellulose, 29% xylan (hemicellulose), 19% acid-insoluble lignin, 6% acid-3 soluble lignin, 6% ash, and 3% protein [16]. 4 Separate simulations were developed for both SHF and SSF, with pretreatment using AMD, 5 and for comparison, H2SO4. There were thus four scenarios considered, i.e. AMD pre-6 treatment with SHF (A-SHF), H2SO4 pre-treatment with SHF (H-SHF), AMD pre-treatment 7 with SSF(A-SSF), and H2SO4 pre-treatment with SSF (H-SSF). 8 The model made use of the NREL database for chemical compounds and structural 9 components in lignocellulosic biomass that were used in the DW1102A Aspen Plus® 10 simulation [21]. This study made use of the NRTL equations of state for modeling of vapor-11 liquid equilibrium. 12 2.1.2 Process description 13 The process consists of three sections: pre-treatment (Figure 1), hydrolysis and 14 fermentation (either SHF (Figure 2) or SSF (Figure 3)) and product separation (Figure 4). 15 6 1 Figure 1: Layout of the pre-treatment section used in the Aspen Plus® simulation 2 3 Figure 2: Layout of the SHF section used in the Aspen Plus® simulation 4 5 Figure 3: Layout of the SSF section used in the Aspen Plus® simulation 6 7 Figure 4: Layout of the product separation section used in the Aspen Plus® simulation 8 2.1.2.1 Pre-treatment 9 In the pre-treatment section, grass is fed into a semi-batch reactor (PRETREAT) where it 10 undergoes pre-treatment by either AMD or H2SO4 which has been heated to 90°C (PRE-HX1 11 & PRE-HX2). Note: AMD and H2SO4 are not used in the same process but in different 12 7 scenarios for comparison. In pre-treatment, xylan is hydrolyzed to xylose, breaking up the 1 ligno-hemicellulosic structure, making cellulose more susceptible to enzymatic hydrolysis. 2 This was represented in Aspen Plus® through the hydrolysis of xylan (C5H8O4) to xylose 3 (C5H10O5). The pre-treated biomass is separated from the liquid (PREFILT) and sent to the 4 hydrolysis and fermentation section. AMD or H2SO4 flow through the grass in the reactor 5 continuously, while the grass is replenished at various intervals. Multiple pretreatment 6 reactors are thus required to allow for the continuous operation of the process during 7 loading/ removal of grass. The liquid product is first used to heat the incoming liquid stream 8 before going to further treatment/ disposal (not considered in this study). 9 The flowrate of AMD/H2SO4 was estimated assuming that the ratio of biomass: AMD/H2SO4 10 consumed by the process was 0.2 as determined as optimal by a previous study [16]. The 11 volume of pure H2SO4 that was required was then calculated assuming a 0.5% wt/wt H2SO4 12 concentration. A summary of the blocks used in the simulation and their specifications is 13 given in Table 1. 14 2.1.2.2 SHF 15 In SHF, pre-treated biomass is fed into the batch hydrolysis the reactor (HYDROLYS) which is 16 operated at 50°C. Make up water and water recycled from product separation section and 17 cooled to 50°C, is fed into this reactor, along with cellulase enzymes, to achieve a solid 18 loading of 20 wt% biomass. Cellulase enzymes catalyze the hydrolysis of cellulose to 19 glucose. After hydrolysis is complete, the liquid fraction of the product stream is separated 20 from the solid fraction (FILTER), cooled to 35°C (FERMCHIL) and fed into the fermenter 21 (FERMENT). An inoculum of Saccharomyces cerevisiae is fed into the fermenter. 22 Saccharomyces cerevisiae facilitates the conversion of glucose to ethanol and CO2. Gas-23 8 phase CO2 produced in fermentation will leave the top of the fermenter. Solid biomass is 1 then separated from the liquid product stream (FILTER) which goes to the product 2 separation section. 3 A summary of the blocks used in the simulation and their specifications is given in Table 1. 4 2.1.2.3 SSF 5 In SSF, pre-treated biomass is fed into the batch SSF reactor which is operated at 38°C. 6 Make-up water and water recycled from the product separation section are blended and 7 cooled to 38°C before passing into the SSF reactor, together with cellulase enzymes and 8 Saccharomyces cerevisiae, to achieve a biomass solid loading of 20 wt%. In this reactor, 9 there is the simultaneous hydrolysis of cellulose to glucose and fermentation of glucose to 10 ethanol. Gas-phase CO2 produced in fermentation will leave the top of the fermenter. Solid 11 biomass is then separated from the liquid product stream (FILTER) which goes to the 12 product separation section. 13 A summary of the blocks used in the simulation and their specifications is given in Table 1. 14 2.1.2.4 Product separation 15 The product separation section consists of two distillation columns. In the 1st distillation 16 column (BEERCOL) the distillate stream contains 40% ethanol, with a 95% recovery of 17 ethanol to the distillate. The second column (RECTCOL) produces a distillate of 90% ethanol 18 with a recovery of 95% ethanol to the distillate. The bottoms stream from both columns is 19 recycled/ purged as necessary. 20 A summary of the blocks used in the simulation and their specifications is given in Table 1. 21 2.1.3 Simulation specifications 22 9 All reactors were modeled using an RSTOICH block with conversions determined in previous 1 experiments [16]. In SSF the hydrolysis and fermentation sections were modeled using 2 separate blocks for hydrolysis (HDROLYSIS) and fermentation (FERMNET) although this 3 process would occur in one reactor. This was done to increase model understanding and 4 troubleshooting capabilities and is valid as the hydrolysis and fermentation reactions would 5 have occurred in series anyway if they were in the same RSTOICH block. In both SHF and SSF 6 a FLASH block (CO2FLASH) has been included after the FERMENT block to model the gas-7 liquid equilibrium of CO2 and ethanol entering the product separation section. 8 Liquid streams entering/exiting reactors were heated/cooled to appropriate temperatures, 9 by heat exchangers modeled by the HX block (2 stream heat exchanger) or HEATER block 10 (heat exchanger with a utility) as required. The solid-liquid separation was modeled using 11 SSPLIT block. Distillation columns were modeled with DSTWU block, which uses the Winn-12 Underwood-Gilliland shortcut method to model distillation columns based on the 13 assumptions of constant relative volatilities and constant molar overflow. 14 The presence of CO2 made the modeling of distillation with DSTWU blocks technically 15 difficult, and hence it was assumed that there was no CO2 present. This was accomplished 16 by the incorporation of a SEP block (CO2REMOV) prior to distillation. 17 A summary of all block design specifications is presented in Table 1. A summary of the 18 design specifications that were manipulated by Aspen Plus® is presented in Table 2. 19 Table 1: Summary of block specifications used in Aspen Plus® simulations 20 Block Name Block Type Specification P re -t re at m en t FEEDCHILL HEATER Thot, out = 50°C PRE- HX1 HX Thot, out = 30°C PRE- HX2 HEATER Tcold, out = 90°C PRETREAT RSTOICH RX1: Xylan → Xylose =0.5 PREFILT SSPLIT 1-50 = 100% solids 1-41 = 90 % liquid 10 SH F HYDROLYSIS RSTOICH Time = 168 Hrs RX1: Cellulose → Glucose FILTER SSPLIT 2-50 = 100% solid 2-60= 95% liquids FERMCHILL HEATER Tcold, out = 35°C FERMENT RSTOICH RX1: Glucose → 2CO2 + 2 EtOH X=0.90 B1 FLASH T=35°C SS F FEEDCHILL HEATER Thot, out = 38 HYDROLYSIS RSTOICH time = 168 Hrs RX1: Cellulose → Glucose FERMENT RSTOICH RX1: Glucose → 2CO2 + 2 EtOH X=0.90 CO2FLASH FLASH Tcold, out = 38°C FILTER SSPLIT 2-90 = 100% solid 2-80= 95% liquids SE P CO2REMOV SEP 3-10 = 100% CO2 BEERCOL DSTWU RRATIO = 1.1 times minimum RECTCOL DSTWU RRATIO = 1.1 times minimum Table 2: Summary of design specifications used in Aspen Plus® simulation 1 Design Specification Specification Adjusted variable DS1 Reaction time = 168 hrs Glucose conversion in HYDROLYS block DS2 40% ethanol in BEERCOL distillate Recovery of water to RECTCOL distillate DS3 90% ethanol in RECTCOL distillate Recovery of water RECTCOL to distillate DS4 Flow rate of water in reactor feed = 20000 kg/hr Flow rate of make-up water to FEEDMIX 2 In hydrolysis, the conversion of cellulose to glucose, and conversion of glucose to ethanol, 3 was modeled using kinetics determined in a previous study [16] (Table 3). The conversion of 4 cellulose to glucose was varied using a design specification on the total reaction time. 5 Table 3: Summary of empirical equations 6 Reaction Equation SHF Hydrolysis G = (m1,MW + C1,M)ln(t) + (m1,CW + C1,C) SHF Fermentation G = G0 – rg  tf G = G0 – rlg  6 -rs  (tf – 6 ) for tf  0:6 hr for tf  6:12 hr SSF E = (m2,MW + C2,M)ln(t) + (m2,Cln(W) + C2,C) Where G is the glucose concentration (g/L), W is the solid loading fraction of biomass (wt%) th, tf, tssf are the hydrolysis, fermentation and SSF time respectively (hrs), G0 is the initial glucose concentration at the start of fermentation (g/L), rg and rs are the rates of fermentation in different phases of fermentation (g/L/hr), E is the ethanol concentration in SSF (g/L), and mi,j and ci,j are empirically determined constants [16] . 7 Pre-treatment time required for AMD pre-treatment (3 days) and H2SO4 pre-treatment (1 8 day) was based on previous studies [16]. 9 11 2.2 Operating Costs 1 Estimation of operating costs was performed using the Aspen Plus® inbuilt utility and 2 operating cost tools. This was done by defining three utilities and then assigning them to 3 certain operating blocks. The three utilities that were defined are low-pressure steam, 4 cooling water and refrigerated water (Table 4). Low-pressure steam, cooling water, and 5 refrigerated water were chosen as they are suitable for the heating of incoming 6 AMD/H2SO4, the condensation of distillate streams, and the cooling of hydrolysate from the 7 hydrolysis reactor (50°C) before it enters the fermenter (35°C), respectively. 8 The cost of the utilities was estimated based on the method of Ulrich and Vasudevan (2006), 9 assuming that steam was produced in a natural gas boiler and the electricity used in the 10 production of cooling and refrigeration water was produced in a coal-based power plant. 11 The cost of fuel was assumed to be 3.57 USD/GJ for natural gas and 0.95 USD/GJ for 12 coal[32]. 13 Table 4: Summary of the utilities used in the Aspen Plus® process simulation 14 Low-pressure steam Cooling water Refrigerated water Cost (USD/GJ) 6.15 1.06 5.08 Tin 125 30 20 Tout 124 45 30 Enthalpy change kJ/kg 2191 62.59 41.74 Used at All reboilers All condensers, FEEDCHILL (SHF) FERMCHILL (SHF) FEEDCHILL (SSF) Pressure (atm) 2.29 1 1 Vapor fraction inlet/ outlet 1/0 0/0 0/0 15 The DSTWU block does not allow for the allocation of utilities. The utility requirements were 16 thus calculated manually and added to utilities calculated in Aspen Plus®. 17 The cost of the grass was taken to 35 USD/ton, as this represents the lowest feedstock cost 18 found in literature. This was based investigation into the market value of the feedstock. 19 12 Due to the lack of physical property data enzyme were not included in the Aspen Plus® 1 simulation, however, the cost of enzyme was taken into account. It was assumed enzyme 2 was dosed at 0.20 mg/g pre-treated biomass based on previous studies [33]. It was assumed 3 that the cost of enzyme was (1.25 USD/kg enzyme). This is once again the cheapest enzyme 4 cost found in literature [34]. The cost of H2SO4 was taken to be 80 USD/ton [35]. 5 The market value of the product was taken to be 815.00 USD/ton [2] 6 2.3 Capital Costs 7 The size of all heat exchangers, columns and pumps were taken from the Aspen Plus® 8 simulation. Fortran script was written in calculator blocks to evaluate the volumes of all 9 reactors (PRETREAT, HYDROLYSIS, FERMENT) based on the process flow rates and kinetics of 10 reactions determined previously [16] (Table 3). All equations used for the determination of 11 reactor volumes are presented in the supplementary data. 12 An estimation of the capital costs for individual process units was performed using the 13 Capcost2017 Program (www.richardturton.faculty.wvu.edu). A summary of the equipment 14 accounted for and its estimated cost is presented in the supplementary data. The cost of the 15 (hydrolysis, fermentation, and pre-treatment reactors) was estimated using prices from the 16 NREL 2011 [21] design report adjusted to 2017 prices using the Chemical Engineering Price 17 Cost Index (CEPCI). Due to the PRETREAT reactor having similar solid loading, batch time and 18 temperatures to the HYDROLYSIS reactor, it was assumed that the price per volume of these 19 reactors would be the same. The total installed capital cost was estimated from the total 20 equipment cost using the Lang factor method with a Lang factor of 4.74 [36]. 21 2.4 Economic Evaluation 22 http://www.richardturton.faculty.wvu.edu/ 13 A simple economic evaluation in which the non-discounted payback period was determined 1 for the process based on the estimated capital costs and annual profit. This evaluation only 2 considered the operating costs mentioned in the operating costs section but did not 3 consider other operating costs such as salaries or tax. 4 3 Results and Discussion 5 3.1 Process flow sheet 6 A full stream table of all streams can be found in the supplementary data. 7 The feedstock requirements for scenarios with the same pre-treatment are the same (Table 8 5). This is due to all scenarios have the same basis (90 tons/hr grass) as well as having the 9 same mass balance around the pre-treatment process section, resulting in the same 10 quantity of biomass entering the hydrolysis and fermentation section and hence the same 11 enzyme usage. Scenarios that use H2SO4 pre-treatment also require H2SO4. 12 Table 5: Summary of the raw material costs 13 Raw material unit value H2SO4 Pre-treatment AMD pre-treatment USD/kg ton/yr MUSD/yr ton/yr MUSD/yr Grass 0.035 750 000 25.5 750 000 25.5 Enzyme 1.25 5 500 6.93 5500 6.93 H2SO4 0.08 18 000 1.50 0 0 Total 33.9 32.4 14 The amount of ethanol produced in different scenarios varied substantially ranging from 15 38 000 to 60 000 ton/year (Table 6). Scenarios with H2SO4 pre-treatment produced more 16 ethanol than the scenarios with AMD pre-treatment. The scenarios with SHF reactor 17 configuration produced more ethanol than the scenarios with SSF reactor configuration. The 18 higher final ethanol volumes (in SHF compared to SSF and H2SO4 compared to AMD) are due 19 14 to higher concentrations of ethanol achieved in fermentation. This reflects the experimental 1 results on which the reaction models used in this study are based (Table 3) [16]. 2 3 Table 6: Summary of ethanol yield, volume produced and revenue from sales for different 4 scenarios. 5 Ethanol Yield Ethanol Volume Revenue Scenario* L/dry ton ton/Year MUSD/yr A-SHF 58.4 44 000 35.7 H-SHF 80.7 60 000 49.3 A-SSF 50.3 38 000 30.7 H-SSF 53.7 40 000 32.8 * A = AMD; H = H2SO46 7 The ethanol yield ranged from 58.4 – 80.7 L/dry ton. This is low in comparison to other 8 techno-economic evaluations that report yields of up to 341 L/dry ton [37]. One explanation 9 of the comparatively low yield is that other studies are fermenting both C5 and C6 sugars 10 [21], whereas in this study the C5 sugars are required as a carbon source for the SRB. 11 Another explanation is that the yields in this study are based on experiments that were 12 conducted without access to tailored cellulase mixtures, advanced bioreactors or genetically 13 modified fermentative organisms [16]. 14 The heating and cooling duty vary with each scenario from 2 340 to 2 830 TJ/yr (Table 7). 15 Total duties required are higher for H2SO4 pre-treatment than AMD pre-treatment and 16 higher for SHF reactor configuration than SSF reactor configuration. The variation in the 17 heating/cooling duties can mainly be attributed to the different reflux/reboiler 18 requirements in each scenario. 19 Table 7: Summary of utility duties and utility costs for different scenarios. 20 Utility Duty Utility Cost Scenario* TJ/yr MUSD/yr A-SHF 2 690 11.4 15 H-SHF 2 830 11.9 A-SSF 2 340 9.06 H-SSF 2 360 9.13 * A = AMD; H = H2SO41 2 3.2 Operating costs and revenue 3 As can be seen in Table 5 the feedstock costs are the same for scenarios with the same pre-4 treatments (i.e. A-SHF & A-SSF; H-SHF & H-SSF). Scenarios with AMD pre-treatment do not 5 require H2SO4 and hence have a lower raw material cost. Although there are many other 6 feedstocks, only grass, cellulase enzyme, and H2SO4 were accounted for as these are 7 considered to be the most significant costs. 8 The variation in ethanol revenue (30.7 MUSD/yr – 49.3 MUSD/yr) follows the same trends 9 as the variation in ethanol produced (Table 6). 10 The trends in the utility costs follow the trend in utility duty (Table 7), with H-SHF 11 (11.9 MUSD/yr) being the highest, followed by A-SHF (11.4 MUSD/yr) followed by H-SSF 12 (9.13 MUSD/yr) and lastly A-SSF (9.06 MUSD/yr). 13 For all scenarios except for H-SHF, the operating costs are higher than the revenue and 14 hence not profitable (Figure 5). This is mainly due to low ethanol yield and hence revenue. 15 Although scenario H-SHF makes a profit, it is small (3.5 MUSD/yr). Considering that this 16 evaluation assumes the cheapest raw material costs, favorable ethanol price, and ignores 17 various other operating costs (salaries, tax, etc.), it is extremely unlikely that this scenario is 18 feasible. The low revenue achieved in each scenario is due to the product yields achieved in 19 hydrolysis and fermentation. 20 16 1 Figure 5: Summary of operating costs, revenue, and profit for different scenarios (A = 2 AMD; H = H2SO4) 3 3.3 Capital Costs 4 A summary of the estimated capital costs for different scenarios is presented in Figure 6. 5 6 Figure 6: Summary of the capital costs for each processes section, and installed plant cost 7 (accounting for purchase and installation all major and ancillary equipment) for the 8 different scenarios (A = AMD; H = H2SO4) 9 As can be seen in Figure 6 (Table 8) the capital cost for the product separation section is 10 substantially less than the capital cost for pre-treatment and hydrolysis and fermentation 11 sections. The large capital cost of the pre-treatment and hydrolysis and fermentation 12 sections is due to the large reactor volumes required for both pre-treatment and hydrolysis. 13 The capital cost for pre-treatment is approximately the same for scenarios with the same 14 35.7 49.3 30.7 32.8 -32.42 -33.91 -32.42 -33.91 -11.4 -11.9 -9.1 -9.1 -8.1 3.5 -10.7 -10.2 -60 -40 -20 0 20 40 60 A-SHF H-SHF A-SSF H-SSF Co st s / R ev en ue / Pr of it (M U SD /y ea r) Revenue Utilities Raw Materials Profit -43.8 -45.8 -41.5 -43.0 37.3 16.9 37.3 16.9 41.6 41.6 55.5 55.5 1.03 0.75 0.94 0.95 299 222 350 273 0 100 200 300 400 500 A-SHF H-SHF A-SSF H-SSF Co st (M U SD ) Pretreatment Hydrolysis & Fermentation Product Separation Installed Plant cost 379 281 444 345 17 type of pre-treatment i.e. AMD or H2SO4. The capital cost for hydrolysis and fermentation is 1 approximately the same for scenarios with the same hydrolysis and fermentation 2 configuration i.e. SSF or SHF. 3 The capital cost for scenarios with H2SO4 pre-treatment was found to be lower than 4 scenarios with AMD pre-treatment. This can be attributed to a lower capital cost of the pre-5 treatment area, for scenarios with H2SO4 pretreatment, compared with scenarios with AMD 6 pre-treatment. Scenarios with H2SO4 pre-treatment have a shorter pre-treatment time (1 7 day) than scenarios with AMD pretreatment (3.5 days), to achieve the desired level of pre-8 treatment. The shorter pre-treatment time results in smaller pre-treatment reactor 9 volumes, which are less expensive. 10 The capital cost for scenarios with SHF was found to be lower than scenarios with SSF. This 11 is due to the hydrolysis reactor having a simpler design, and hence less expensive, than the 12 SHF reactor. Although the SHF scenarios require an additional fermentation reactor, the 13 cost of this is small in comparison to the hydrolysis and SHF reactors due to a significantly 14 shorter batch time, and hence volume. 15 The installed plant cost (including capital cost, transportation, and installation of the entire 16 process) was estimated using the Lang Factor method. This method assumes that the total 17 installed plant cost can be estimated by multiplying the capital cost of the major process 18 units by a Lang factor. This is inclusive of the capital cost of all major and ancillary 19 equipment (piping, electrical, control, etc.); transportation and installation of major and 20 ancillary equipment; required civil works; and engineering services. A Lang factor of 4.74 21 was assumed for this study [36]. The total installed plat cost ranges from 281 MUSD to 22 444MUSD. 23 18 A previous techno-economic evaluation [21] determined a total installed plant cost of 232 1 MUSD, which is the same order of magnitude. The study by Humbird et al. (2011) [21] 2 included various process sections such as enzyme production and wastewater remediation 3 that were not included in this evaluation. The total installed plant cost determined in this 4 evaluation is thus significantly higher, even if compared to an inflation-adjusted value. This 5 high installed plant cost can be attributed to the large reactor volumes required for both 6 pre-treatment and hydrolysis, resulting in a large capital cost for these sections. 7 3.4 Economic Evaluation 8 Table 8: Summary of operating costs, capital cost and payback period 9 Scenario Name* Capital Cost Operating Cost Revenue Profit Payback Period MUSD MUSD/yr MUSD/yr MUSD/yr yr A-SHF 378.8 43.83 35.69 -8.14 - H-SHF 281.0 45.82 49.30 3.48 80.7 A-SSF 444.1 41.47 30.72 -10.75 - H-SSF 345.3 43.04 32.83 -10.21 - * A = AMD; H = H2SO4 10 The payback period was only calculated for scenario H-SHF, as this was the only scenario 11 that makes a profit (Table 8). 12 The payback period for Scenario H-SHF is extremely long (80.7 years), indicating that this 13 process is not economically feasible. Considering that this evaluation assumed favorable 14 economic conditions and did not take into account various operating costs (tax, salaries) this 15 scenario would be of even less feasibility. The major issue making this process unfeasible is 16 the low yield of ethanol from feedstock. This results in a low revenue as well as a high 17 capital cost as the enzymatic hydrolysis time has been extended to try to achieve higher 18 product yield. 19 To reduce the capital cost and hence payback period all scenarios were analyzed using a 20 hydrolysis/ SSF reaction time of 84 hours as opposed to 168 hours. This resulted in a slightly 21 19 lower volume of ethanol produced, and a significantly lower capital cost for the hydrolysis 1 and fermentation section. Although this reduces the capital cost for all scenarios, the 2 decreased ethanol production, and hence revenue, resulted in all scenarios, including H-SHF 3 making a loss. 4 Although this evaluation indicates that all scenarios are currently unfeasible, some insightful 5 observations can be made. The SHF reactor configuration is significantly cheaper and 6 produces more ethanol than the SSF configurations. This is true for both pre-treatment with 7 AMD and pre-treatment with H2SO4. The SHF reactor configuration also allows for the 8 production of various products through different biochemical transformations of the 9 glucose intermediary. The production of small quantities of high-value chemicals, in addition 10 to large quantities of biofuels, can contribute towards the economic feasibility of a process 11 [12]. It is thus recommended that all future studies into the production of bioethanol using 12 AMD pre-treated biomass should focus on the SHF reactor configuration. 13 Incorporating the process of bioethanol production into a process that simultaneously 14 remediates AMD could also improve its feasibility. After pre-treatment AMD contains xylose 15 released during pre-treatment, which could be utilized as a carbon source for sulfate-16 reducing bacteria to remediate the AMD. The integration of the two processes would be 17 financially beneficial to both as the cost of pre-treatment of the biomass could be shared 18 with both processes. As the remediation of AMD already has a cost associated with it, any 19 savings on this could be attributed as a profit stream for the integrated process. It is thus 20 recommended that a techno-economic evaluation of this envisioned integrated bio-ethanol 21 production/AMD remediation process be conducted. 22 20 As the product yield was found to be low, it is recommended that further experimental 1 work be carried to determine the optimal conditions of enzymatic hydrolysis in the SHF 2 configuration. Tailored enzyme mixtures should be investigated in conjunction with 3 hydrolysis operating conditions such as temperature, pH and mixing. 4 4 Conclusion 5 This techno-economic evaluation found scenarios A-SHF, A-SSF and H-SHF to be 6 unprofitable, while scenario H-SHF was only marginally profitable (3.5 MUSD/year). 7 Although the H-SHF scenario was profitable, the payback period calculated was too long to 8 be feasible (80.5 years). Although all scenarios are unfeasible SHF was shown to be more 9 favorable than SSF. 10 Due to a lack of previous quantitative research, this study was limited to data from only one 11 previous study, which was both feedstock (Weeping lovegrass) and product (ethanol) 12 specific and resulted in a low product yield (58.4 – 80.7 L/dry ton) [16]. The low product 13 yield decreased the feasibility of the process, especially considering that ethanol is a low 14 value biochemical. 15 Optimization of enzymatic hydrolysis, through improved enzymes and operating conditions, 16 could substantially increase product yield and improve the economic feasibility of the 17 process. The economic feasibility could also be substantially improved if it were to be 18 coupled with an AMD remediation process. AMD remediation has a cost associated with its 19 so any savings could be considered as a source of income. 20 Although this process is not currently feasible, further research and process development 21 are encouraged. This process has the potential to produce environmentally friendly 22 21 alternatives to fossil fuel products, while simultaneously remediating AMD. This could be 1 highly beneficial for a developing country such as South Africa, as it will help protect the 2 environment and increase both energy and water security without impacting food 3 production. 4 5 Acknowledgments 5 This work is based on research supported in part by the National Research Foundation of 6 South Africa (Grant Numbers: 111864). The support of the University of the Witwatersrand, 7 Johannesburg is also acknowledged. 8 6 References 9 [1] IPCC, Climate Change 2014: Contribution of Working Groups I, II and III to the Fifth10 Assessment Report of the Intergovernmental Panel on Climate Change, in: Core11 Writing Team, R.K. Pachauri, L.A. Meyer (Eds.), IPCC, Geneva, Switzerland, 2014: p.12 151.13 [2] E4TECH, RE-CORD, WUR, From the Sugar Platform to biofuels and biochemicals, 2015.14 doi:contract No. ENER/C2/423-2012/SI2.673791.15 [3] J. Lange, Lignocellulose conversion : an introduction to chemistry , Biofuels, Bioprod.16 Biorefining. 1 (2007) 39–48. doi:10.1002/bbb.17 [4] E. Searcy, P.C. Flynn, Processing of Straw / Corn Stover : Comparison of Life Cycle18 Emissions, Int. J. Green Energy. 5 (2008) 423–437. doi:10.1080/15435070802498010.19 [5] N. Mosier, C. Wyman, B. Dale, R. Elander, Y.Y. Lee, M. Holtzapple, M. Ladisch,20 Features of promising technologies for pretreatment of lignocellulosic biomass,21 Bioresour. Technol. 96 (2005) 673–686. doi:10.1016/j.biortech.2004.06.025.22 [6] Z. Yu, H. Zhang, Ethanol fermentation of acid-hydrolyzed cellulosic pyrolysate with23 Saccharomyces cerevisiae, Bioresour. Technol. 90 (2003) 95–100. doi:10.1016/S0960-24 8524(03)00093-2.25 [7] P. Alvira, E. Tomas-Pejo, M. Ballesteros, M.J. Negro, Pretreatment technologies for an26 efficient bioethanol production process based on enzymatic hydrolysis: A review,27 Bioresour. Technol. 101 (2010) 4851–4861. doi:10.1016/j.biortech.2009.11.093.28 [8] D.P. Maurya, A. Singla, S. Negi, An overview of key pretreatment processes for29 biological conversion of lignocellulosic biomass to bioethanol, 3 Biotech. 5 (2015)30 597–609. doi:10.1007/s13205-015-0279-4.31 22 [9] M.J. Taherzadeh, K. Karimi, Pretreatment of lignocellulosic wastes to improve ethanol1 and biogas production: A review, Int. J. Mol. Sci. 9 (2008) 1621–1651.2 doi:10.3390/ijms9091621.3 [10] N.W. Burman, C. Sheridan, L. van Dyk, K.G. Harding, Modelling of low temperature4 dilute sulfuric acid pre-treatment of South African grass, Bioresour. Technol. Reports.5 4 (2018) 21–28. doi:10.1016/j.biteb.2018.08.014.6 [11] E. Magowo, K. Rumbold, C. Sheridan, The Utilization of Cellulosic Biomass in treating7 AMD and the Subsequent Generation of Fermentable Sugars, in: 10th Int. Conf. Acid8 Rock Drain. IMWA Anu. Conf., 2015: pp. 1–12.9 [12] N.W. Burman, K.G. Harding, C.M. Sheridan, L. Van Dyk, Evaluation of a combined10 lignocellulosic / waste water bio-refinery for the simultaneous production of valuable11 biochemical products and the remediation of acid mine drainage, Biofuels, Bioprod.12 Biorefining. 12 (2018) 649–664. doi:10.1002/bbb.1880.13 [13] D.B. Johnson, K.B. Hallberg, Acid mine drainage remediation options: A review, Sci.14 Total Environ. 338 (2005) 3–14. doi:10.1016/j.scitotenv.2004.09.002.15 [14] D.B. Johnson, K.B. Hallberg, Biogeochemistry of the compost bioreactor components16 of a composite acid mine drainage passive remediation system, Sci. Total Environ. 33817 (2005) 81–93. doi:10.1016/j.scitotenv.2004.09.008.18 [15] G.S. Simate, S. Ndlovu, Acid mine drainage: Challenges and opportunities, J. Environ.19 Chem. Eng. 2 (2014) 1785–1803. doi:10.1016/j.jece.2014.07.021.20 [16] N.W. Burman, K.G. Harding, C.M. Sheridan, Lignocellulosic bioethanol production21 from grasses pre-treated with acid mine drainage: Modeling and comparison of SHF22 and SSF Author links open overlay panel, Bioresour. Technol. Reports. 7 (2019).23 doi:10.1016/j.biteb.2019.100299.24 [17] B. Ramla, C. Sheridan, The potential utilisation of indigenous South African grasses for25 acid mine drainage remediation, Water SA. 41 (2015) 247–252.26 doi:10.4314/wsa.v41i2.10.27 [18] N.W. Burman, L. van Dyk, F. Postma, C. Sheridan, G.S. Simate, K.G. HArding, Flow28 Sheet and Sensitivity Analyses for the Bio-Remediation of Acid Mine Drainage Using29 Sulfate Reducing Bacteria and South African Grasses, in: 2nd Int. Conf. Energy,30 Evironment Clim. Chang., Pointe Aux Piments, Mauritius, 2017.31 [19] D.K. Westensee, K. Rumbold, K.G. Harding, C.M. Sheridan, L.D. van Dyk, G.S. Simate,32 F. Postma, The availability of second generation feedstocks for the treatment of acid33 mine drainage and to improve South Africa’s bio-based economy, Sci. Total Environ.34 637–638 (2018) 132–136. doi:10.1016/j.scitotenv.2018.04.410.35 [20] E. Gnansounou, A. Dauriat, Bioresource Technology Techno-economic analysis of36 lignocellulosic ethanol : A review, Bioresour. Technol. 101 (2010) 4980–4991.37 doi:10.1016/j.biortech.2010.02.009.38 [21] D. Humbird, R. Davis, L. Tao, C. Kinchin, D. Hsu, A. Aden, P. Schoen, J. Lukas, B. Olthof,39 M. Worley, D. Sexton, D. Dudgeon, Process Design and Economics for Biochemical40 Conversion of Lignocellulosic Biomass to Ethanol, Golden, Co., 2011.41 http://www.nrel.gov/docs/fy11osti/47764.pdf.42 23 [22] F.K. Kazi, J.A. Fortman, R.P. Anex, D.D. Hsu, A. Aden, A. Dutta, G. Kothandaraman,1 Techno-economic comparison of process technologies for biochemical ethanol2 production from corn stover q, Fuel. 89 (2010) S20–S28.3 doi:10.1016/j.fuel.2010.01.001.4 [23] E. Sendich, M. Laser, S. Kim, H. Alizadeh, L. Laureano-perez, B. Dale, L. Lynd, Recent5 process improvements for the ammonia fiber expansion ( AFEX ) process and6 resulting reductions in minimum ethanol selling price, Bioresour. Technol. 99 (2008)7 8429–8435. doi:10.1016/j.biortech.2008.02.059.8 [24] B. Bals, C. Wedding, V. Balan, E. Sendich, B. Dale, Evaluating the impact of ammonia9 fiber expansion ( AFEX ) pretreatment conditions on the cost of ethanol production,10 Bioresour. Technol. 102 (2011) 1277–1283. doi:10.1016/j.biortech.2010.08.058.11 [25] D. Klein-marcuschamer, P. Oleskowicz-popiel, B.A. Simmons, H.W. Blanch,12 Technoeconomic analysis of biofuels : A wiki-based platform for lignocellulosic13 biorefineries, Biomass and Bioenergy. 34 (2010) 1914–1921.14 doi:10.1016/j.biombioe.2010.07.033.15 [26] H. Huang, S. Ramaswamy, W. Al-dajani, U. Tschirner, R.A. Cairncross, Effect of16 biomass species and plant size on cellulosic ethanol : A comparative process and17 economic analysis, Biomass and Bioenergy. 33 (2009) 234–246.18 doi:10.1016/j.biombioe.2008.05.007.19 [27] M. Laser, E. Larson, B.E. Dale, M. Wang, N. Greene, L.R. Lynd, Comparative analysis of20 efficiency , environmental impact , and process economics for mature biomass refi21 ning scenarios, Biofuels, Bioprod. Biorefining. 3 (2009) 247–270. doi:10.1002/bbb.22 [28] C. Piccolo, F. Bezzo, A techno-economic comparison between two technologies for23 bioethanol production from lignocellulose, Biomass and Bioenergy. 33 (2009) 478–24 491. doi:10.1016/j.biombioe.2008.08.008.25 [29] National Accademy of Sciences, National Academy of Engineering, National Research26 Council, The national academies press, 1st ed., The National Academies Press,27 Wachington, D.C., 2009. doi:10.17226/12620.28 [30] C.L.L. Robus, L.D. Gottumukkala, E. van Rensburg, J.G. Görgens, Feasible process29 development and techno-economic evaluation of paper sludge to bioethanol30 conversion : South African paper mills scenario, 92 (2016) 333–345.31 doi:10.1016/j.renene.2016.02.017.32 [31] G.D. Ulrich, P.T. Vasudevan, How to Estimate Utility Costs for a number of utilities,33 Chem. Eng. (2006) 66–69.34 [32] Department of Energy, 2018 Energy Price Report, Pretoria, 2018.35 http://www.energy.gov.za/files/media/explained/Energy-Price-Report-2018.pdf.36 [33] S. Ellilä, L. Fonseca, C. Uchima, J. Cota, G.H. Goldman, M. Saloheimo, V. Sacon, M.37 Siika-Aho, Development of a low-cost cellulase production process using Trichoderma38 reesei for Brazilian biorefineries, Biotechnol. Biofuels. 10 (2017) 1–17.39 doi:10.1186/s13068-017-0717-0.40 [34] G. Liu, J. Zhang, J. Bao, Cost evaluation of cellulase enzyme for industrial-scale41 cellulosic ethanol production based on rigorous Aspen Plus modeling Cost evaluation42 24 of cellulase enzyme for industrial-scale cellulosic ethanol production based on 1 rigorous Aspen Plus modeling, Bioprocess Biosyst. Eng. 39 (2016) 133–140. 2 doi:10.1007/s00449-015-1497-1. 3 [35] Argus Media, Argus Sulfuric Acid, London, 2018. https://www.argusmedia.com/-4 /media/Files/sample-reports/argus-sulphuric-5 acid.ashx?la=en&hash=28D8CDEF485998FED5806750D90A916D3F3DED19.6 [36] R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, Analysis, Synthesis, and Design of7 Chemical Processes, 1st ed., Pretince Hall, Upper Saddle River, 1998.8 [37] A. Aden, M. Ruth, K. Ibsen, J. Jechura, K. Neeves, J. Sheehan, B. Wallace, L. Montague,9 A. Slayton, J. Lukas, Lignocellulosic Biomass to Ethanol Process Design and Economics10 Utilizing Co-Current Dilute Acid Prehydrolysis and Enzymatic Hydrolysis for Corn11 Stover Lignocellulosic Biomass to Ethanol Process Design and Economics Utilizing Co-12 Current Dilute Acid Prehyd, Golden , Co, 2002.13 14