Microbe and bioprocess performances for sustainable production of biobased 2,3-butanediol in a sugarcane biorefinery; a technoeconomic and environmental analysis

Figure 1 provides an overview of the study methodology, which comprises an extensive sensitivity analysis of the impact of product titre, yield and volumetric productivity on energy and mass flows as well as economic and environmental viability indicators for sugar-based 2,3-BDO production. The statistical FFD was applied to map out economic and environmental merits from bioprocess enhancement within a bioprocess performance of 0.30–0.495 g_2,3-BDO g_sugar⁻¹ for product yield, 1.0–3.0 g_2,3-BDO  L⁻¹ h⁻¹ for volumetric productivity and 50–150 g_2,3-BDO  L^{−1 1} for titre. Each data point prescribed by the FFD was modelled as an Aspen Plus® process flowsheet, with resultant mass & energy balances leveraged in economic assessments and GHG emission calculations.

A 3³ full-factorial design (FFD) having 3 levels (low, medium, high) was adopted for this study translating to 27 separate data points with unique yield-titre-volumetric productivity coordinates, which provided bioprocess descriptions for each process models in Aspen Plus® (see Table 1). The upper limits were based on the maximum theoretical limits for yield, titre and volumetric productivity determined in “Determination of upper and lower limits for the bioprocess performance metrics” section. The lower limits were set arbitrarily to capture all key results obtained from studies on microbial 2,3-BDO production upon thorough literature survey (Ji et al. 2011). These were a yield of 0.30 g_2,3-BDO  gsugar⁻¹, volumetric productivity of 1.0 g_2,3-BDO  L⁻¹ h⁻¹ and titre of 50 g_2,3-BDO  L⁻¹. The medium levels were calculated exactly halfway between the upper and lower limits as per the FFD methodology. Fractional conversions for all scenarios were calculated using appropriate stoichiometric correlations between the substrate and products formed in the metabolic pathway (Section A of Supporting information).

All scenarios were simulated as integrated 2,3-BDO biorefineries annexed to a typical sugar mill of 300 t. h⁻¹ cane crushing capacity. The available A-molasses from the sugar mill (25.4 t. h⁻¹) was diverted as feedstock to the 2,3-BDO biorefinery, eliminating C-molasses production and reducing crystalline sugar output from 38.4 to 27.7 t h⁻¹ (Dogbe et al. 2020). The composition of A-molasses was 22.1% water, 54.4% sucrose and the balance equally shared by glucose and fructose. Omission of further processing of A-molasses in the sugar mill reduces the net energy demand of the sugar mill to allow 1.86 MW of electricity, 15.5 t h⁻¹ high pressure steam (HPS) and 15 t h⁻¹ medium pressure steam (MPS) to be available to use in the biorefinery (Ratshoshi et al. 2021). Details of the combined heat and power plant (CHP) supplying the sugar mill’s energy needs are well documented in a study by Dogbe et al. (2019).

All processes were simulated in Aspen Plus® version 11 (Aspen Technology Inc., Cambridge, MA, USA). Except where otherwise indicated, all chemical inputs were modelled according to one of the existing Aspen Plus® data libraries. Principally, the Electrolyte Non-random Two Liquid (ELECNRTL) and Universal Quasi-Chemical (UNIQUAC) were adopted for modelling vapour–liquid (VLE) and liquid–liquid (LLE) equilibria, respectively. Harvianto et al. (2018) used LLE experimental data to regress binary interactions using the UNIQUAC model with which they were able to extract 2,3-BDO from a fermentation broth using oleyl-alcohol (OA). Binary interaction data from the mentioned study were used to model LLE using the extractor in Aspen Plus®. The distillation column was modelled using the RADFRAC unit while pumps and compressors were modelled as default isentropic components. Lastly, all reactors were simulated on stoichiometric basis as RSTOIC units requiring as a key input, the fractional conversion of all reacting species to the target product.

The schematics of different processing areas for the biorefinery are shown in Fig. 2. A-molasses was sterilized at a temperature of 121 \(^\circ{\rm C}\) to eliminate possible microbial contamination and cooled back to 37 \(^\circ{\rm C}\) for intermittent storage (Ikram-Ul et al. 2004) (see—Fig. 2a). Next, 24.9 t. h⁻¹ of dilution water was added to the A-molasses to bring the feedstock total sugars concentration to 63 brix (630 g L⁻¹) before it was hydrolyzed with invertase to fermentable hexose sugars (fructose and sucrose), assuming a 98% sucrose conversion at a temperature of 60 \(^\circ{\rm C}\) (Białkowska 2016). A portion of the diluted feedstock (about 7 wt%) was subsequently diverted to the seed train section for inoculum cultivation and the balance was sent to the main fermentation area for 2,3-BDO production.

Both fermenters were simulated with operating conditions reported by Jantama et al. (2015) namely, operation temperature and pressure of 37 °C and 1 bar, respectively. NaOH was used for pH regulation and the usage rate was calculated using correlations from Humbird et al. (2011) based on the fermenter working volume for each scenario. Other parameters such as fermentation time, titre, yield, and volumetric productivity were set to conform to the coordinates set by the FFD methodology for each scenario as described in Table 1. Key nutrients for inoculum growth were diammonium phosphate (5.88 kg t⁻¹_A-molasses) and ferric chloride (5.37 kg t⁻¹_A-molasses) with the former being the chief nitrogen source (Jantama et al. 2015). All the nutrients were added to the seed train fermenter during the growth of cells and the effluent from this section was mixed with the remainder of the hydrolyzed feedstock and conveyed to the fermentation section to commence 2,3-BDO production.

After fermentation, the broth was centrifuged (see—Fig. 2b) to remove cells and suspended solids, which were conveyed to the wastewater treatment (WWT) area while the liquid portion was directed to an extraction column for further separation and purification. The extractor column was modelled according to the operating parameters and regressed experimental data for liquid–liquid-extraction of 2,3-BDO from aqueous systems using oleyl alcohol (OA) as solvent (Harvianto et al. 2018). Design specifications in Aspen Plus® were used to vary the number of separation stages between to achieve a target 95% extraction of 2,3-BDO from the aqueous phase (see-Table 2). The product-rich organic phase from the extractor was preheated to 86°C and fed to a vacuum distillation column modelled to operate within actual industrial requirements for a similar process as presented in Harvianto et al. (2018). Here, the molar reflux ratio, number of stages and distillate to feed ratios were the key manipulated variables required to attain a product purity of at least 97 wt% in the distillate stream (see—Table 2). Finally the top stream was fed to an acid digester for removal of acetic acid to attain a final product purity of atleast 98 wt% 2,3-BDO  Lv et al. (2012) reported the option of removing acetic acid from alcohols by employing ion exchange resins or sacrificial alkalis added directly to the final product. In this study, lime was used as the neutralizing agent, which when contacted with acetic acid forms a solid precipitae that would settle to the bottom of the reactor, leaving the clarified supernantant as final product. The acid digester was modelled as a stoichiometric reactor with an assumed 98% fractional conversion of acetic acid to calcium acetate. Here also, the CaO usage was maniputed to achieve a final 2,3-BDO purity of 98wt% for all scenarios. Next, the final product was cooled down to 30°C and sent for storage while the OA rich bottoms stream was recycled to the column extractor.

The water enriched effluents from the fermentation area, separation and purification section, and the biomass from fermentation were all conveyed to the wastewater treatment (WWT) section (see—Fig. 2c). The WWT method used in this study is based on the National Renewable Energy Laboratory (NREL) configuration data (Humbird et al. 2011). Here, the entrained biomass is anaerobically digested to yield biogas, ammonia, and unconverted residue (sludge). The biogas was trapped and stored for subsequent combustion alongside bagasse to meet the process energy demands. In contrast, the sludge and ammonia were further digested aerobically to produce a disposable sludge (solid waste) and a water-rich effluent. The sludge was skimmed off from the bottom of the reactor and disposed of as solid waste while the supernatant was treated using reverse osmosis to yield purified water to be recycled and reused in the biorefinery (Humbird et al. 2011).

The heat demand for the column reboiler was met by a heat transfer fluid system designed according to specification provided by Dowtherm (1997), depicted in Fig. 2d. It operates on a hot oil which is a formulation of distinct aromatics endowed with high thermal stability over a wide operating temperature range (− 35–330 °C). Initially, the hot oil was heated to 350 °C by burning a portion of the available lignocelluloses in a heat chamber and pumped to the reboiler to deliver the required heat demand. On return, the oil was cooled with just enough water to produce a stream rich in utility low-pressure steam (LPS) at 2.2 bar and 124 °C for biorefinery use. Lastly, the hot oil was further cooled down to 100 °C before being recirculated to the heating chamber to restart the cycle.

Energy and mass flow results obtained from the process models developed in Aspen Plus® (as described in Table 1) were used for sizing and costing of the process equipment for each of the 27 scenarios and subsequent determination of capital expenditure (CAPEX), operating expenditure (OPEX) and the minimum selling price MSP. The costing was done using second order capital cost estimation methods based on literature on similar equipment (Woods 2007) and additional biorefinery costing correlations developed by Petersen et al. (2017). A discounted cash flow analysis was conducted for each scenario over a period of 25 years in real terms, assuming a desired discount rate of 20% (real terms) to obtain the MSP, which is the 2,3-BDO at which the net present value is zero under the above conditions. Table 3 summarizes the key economic parameters used for the technoeconomic studies across all scenarios.

To estimate the greenhouse gas emissions for each simulated biorefinery scenario, a cradle to gate analysis (from extraction to dispatch) based on a life-cycle approach was conducted using the Roundtable on Sustainable Biomaterials (RSB) methodology version 4.03 (www.​rsb.​org). The RSB tool is an internationally acknowledged method for approximating the GHG values of processes associated with biomass and biofuels (RSB 2009). The tool divides a typical bioprocessing facility into thee broad areas for GHG estimation purposes namely, feedstock cultivation, transportation and usage, and processing, thus providing GHG emissions for the cradle-to-gate of a sugarcane biorefinery. The RSB tool allows for a maximum of three processing areas. For this study, the sugar mill formed processing area 1 (yielding sugar, A-molasses and bagasse) and the biorefinery was designated processing area 2 (yielding 2,3-BDO). The GHG bequests from all categories are estimated directly using the mass and energy data obtained from each scenario using the inbuilt emission factors allocated to all chemicals used in the process and are summed up to give the gross carbon emission in kg-CO₂eq per unit product or feedstock on either wet or dry basis.

The impact of bioprocess performance on the key energy and mass flows as a function of product titre and yield is shown in Fig. 3. The product yield had the strongest effect on both the total electricity demand and cooling duty compared to the titre and volumetric productivity (see—Fig. 3a, b). Analysis of regression coefficients showed that a 10% improvement in product yield alone reduced electricity and cooling demand per mass unit of 2,3-BDO by 5.2% and 3.7% respectively compared to 1.4% and 2.0% for a 10% step improvement in titre only. For bioprocess performances beyond the mid-range (above 100 g_2,3-BDO  L⁻¹ and 0.40 g_2,3-BDO  g_sugar⁻¹), the energy savings for the same step increments of 10% increased two-fold from product yield and threefold from titre. This is also demonstrated by the increased steepness in the response contours in this region in Fig. 3a, b.

Effect estimates further showed than a 10% incremental change in yield raised the 2,3-BDO output 50 times more than a 10% incremental change in titre within the boundaries of this study (see—Fig. 3c). As a consequence, for a fixed/constant supply of sugar in A-molasses, improving the product yield increased the 2,3-BDO production rate of the biorefinery which generally lowered the unit electricity and cooling requirements per unit mass of product across all scenarios (see—Fig. 3c). The reductions in co-products formation via highly aerobic and exothermic reactions, as a result of technical improvements in the 2,3-BDO yield (see—Section 0), also reduced the requirements for air sparging and chilled water for temperature control in the bioreactors. Thus, the electricity usage from air compression and the total chilling duty, which was included in the overall cooling demand, dropped significantly (see—Fig. 1 of Online Resource).

Conversely, the titre had a stronger influence on total heating requirements per unit mass of 2,3-BDO than product yield (see—Fig. 3d). Analysis of regression coefficients showed that a 10% improvement in titre alone reduced the overall heating demand per kilogram of 2,3-BDO by 3.1% compared to 1.0% for a 10% increase in product yield. The reduced process volumes at high titres reduced the usage rate of process water across the scenarios, which has a high specific heat capacity. This translated to significant energy savings for major energy intensive equipment such as the distillation column and heat exchangers. Also, the heat duties from a-molasses sterilization and hydrolysis sections remained the same for all scenarios, while those for the rest of the process decreased as titre increased; thus, the resulting savings in heating demands were disproportionately large compared to step-changes in titre investigated in the FFD methodology. This is shown by the gradual decrease in the gradient of the response tours in Fig. 3d. Similarly, for process performance above 100 g_2,3-BDO  L⁻¹ and 0. 40 g_2,3-BDO  g_sugar⁻¹, the savings on heating demand from 10% step increments reduced half for improvements in titre but nearly doubled for increments in yield.

Finally, the FFD results showed that volumetric productivity had a negligible effect on process energy requirements per unit mass of product compared to product yield and titre. Hence plots of the effect of yield and titre on mass and energy flows at volumetric productivities of 1.0, 2.0 and 3.0 g_2,3-BDO  L⁻¹ h⁻¹ were identical to those shown in Fig. 3a–d. Increasing the volumetric productivity at fixed product yield and titre meant reducing the fermentation time and vice versa. However, the resultant change in the average process mass and energy flows from altering fermentation time were small particularly when the latter was expressed per unit mass of final product. Overall, the results showed the need to prioritize product yield and titre enhancements over volumetric productivity to achieve meaningful reductions in process energy consumption for 2,3-BDO biorefineries.

The impact of bioprocess performance on the annualized CAPEX and OPEX of the simulated 2,3-BDO scenarios is shown in Fig. 4. Overall, the product yield showed the strongest influence on both economic parameters ahead of titre and volumetric productivity, with the latter having the least influence. Regression analysis showed that a 10% incremental improvement in product yield alone improved the annualized CAPEX and OPEX by 3.0% and 3.9% respectively. Increasing the product yield led to a higher product throughput which translated to economies of scale benefits resulting in a reduced annualized CAPEX (see—Fig. 4a–c) and OPEX (see—Fig. 4d–f) across all scenarios.

Similarly, a 10% improvement in titre alone lowered the annualized CAPEX and OPEX by 1.7% and 1.3% respectively. The reduced total process volumes at increased titres significantly lowered the sizes and cost of key downstream equipment, especially fermenters and digesters for wastewater treatment (see—Fig. 2 of Online Resource). However, the superior economies of scale benefits from product yield improvements overshadowed the CAPEX and OPEX reduction benefits from titre improvements.

Lastly, the effect of a 10% incremental change in volumetric productivity alone was a reduced annualized CAPEX and OPEX by only 0.3% and 0.1% respectively. A high volumetric productivity implies a reduced fermentation time, which lowers the required volume and total capital cost of the bioreactors. However, this did not translate to significant overall reduction in annualized CAPEX because the fermenter cost was dwarfed by the combined cost of the rest of the equipment in the process and the economies of scale benefits from volumetric productivity improvement were insignificant. Thus, enhancement of product yield, and titre should be a priority over volumetric productivity for viable industrial scale production of 2,3-BDO.

It was observed that the gradients of the response contours generally increased with increased bioprocess performance across the scenarios which indicated an increased relative impact of yield over titre particularly in the region above the mid-section of the plots (see—Fig. 4a–f). This implies that step improvements in bioprocess performance in this region, approaching the estimated/theoretical limits in all of the bioprocess performance metrics simultaneously, would lead to particularly larger reductions in annualized CAPEX and OPEX. Analysis of regression coefficients in this region showed that the annualized CAPEX and OPEX reduced twice as much for a 10% improvement in product yield (5.7%), compared to titre (2.9%). The volumetric productivity on the other hand did not show significant impact even in this region.

Figure 5a–f depicts the impact of product yield, titre, and volumetric productivity on the calculated minimum selling price (MSP) for 2,3-BDO across the bioprocess performance space described by the FFD methodology. The MSP was more sensitive to product yield improvements than titre or volumetric productivity, with the latter showing the least impact on the MSP. Regression coefficients analysis showed that a 10% incremental change in product yield alone reduced the MSP of 2,3-BDO by 3.5%, compared to 1.4% and 0.1% reductions in MSP for 10% improvements in titre and volumetric productivity, respectively.

Equally, for bioprocess performance metrics above 0.38 g_2,3-BDO.g_sugar⁻¹, 100 g_2,3-BDO.L⁻¹ and 2.0 g_2,3-BDO  L⁻¹ h⁻¹, the MSP contours become steeper (more vertical than horizontal), which indicates a bioprocess region of greater relative MSP benefits from improvements of product yield over titre (see—Fig. 5a–c) and titre over volumetric productivity (see—Fig. 5d–f). Regression analysis in the same region showed that the MSP was reduced by half in this region though simultaneous 10% incremental improvements in all of the bioprocess metrics. Thus, the reductions in MSPs became more substantial as bioprocess performances approached the theoretical limits, in particular when these were achieved simultaneously. This provided evidence of the need to improve bioprocess performances simultaneously for all performance metrics, rather than focusing on one metric only. With that said, product yield, and titre improvements should be a priority for optimizing the process economics of biobased 2,3-BDO ahead of volumetric productivity. The effect of volumetric productivity on the MSP was over 30 times lower than that of titre and yield and will require a significant step change to offset the economic outcomes obtainable from target product yield and titre values as shown in Fig. 5d–f.

The current best-reported bioprocess performances for microbial 2,3-BDO production from sugars were compared to the predictions from the FFD-based statistical model, to determine whether these were sufficient to achieve an MSP that was equal to or below the current fossil-based price of 1600$ t_2,3-BDO⁻¹ (Huo et al. 2022). From the FFD-based regression model the minimum bioprocess performance required to achieve an MSP of 1600$ t_2,3-BDO⁻¹was estimated to be as 0.44 g_2,3-BDO.g_sugar⁻¹, 127.5 g_2,3-BDO  L⁻¹ and 1.05 g_2,3-BDO.L⁻¹ h⁻¹ shown by the yellow contour on—Fig. 5a, e. This is therefore lowest combination of the three factors required to achieve a commercially competitive process. Bioprocess performance below these numbers for one metric would require a performance upgrade in the others to meet the viability threshold and vice versa. Process viability can be achieved with numerous combination of these metrics; thus, the numbers above could guide experimental studies on microbial 2,3-BDO production with a view of industrial scale production. The best MSP projected from this study was 1355$ t⁻¹, which is 16% lower than the current fossil-based price of 2,3-BDO, and corresponded to product yield, titre and volumetric productivity performances of 0.495 g_2,3-BDO.g_sugar and 150 g_2,3-BDO.L⁻¹ and 3.0 g_2,3-BDO  L.⁻¹ h⁻¹ (point D on Fig. 5c, f). Using the same Aspen Plus® simulations approach, the current state of the 2,3-BSO production technology achieved MSP values in the range of 1434–1659$ t⁻¹, showing economic viability at the minimum acceptable IRR of 20% (real terms) as shown in Table 4. Clearly, strains belonging to S. cerevisiae (HGS37), K. oxytoca (KMSOO5-73 T), K. oxytoca (M1) and K. pneumoniae have the potential to produce 2,3-BDO feasibly in a sugarcane biorefinery using the proposed process configuration as the reported fermentation performance are sufficient to achieve an MSP less than or equal to the current 2,3-BDO fossil-based price of 1600$ t⁻¹_2,3-BDO (see—Table 4).

The cradle to gate lifecycle GHG emission results for the sugar mill section are summarized in Fig. 6 (The total emissions from the sugar mill area were 0.16 kg-CO₂eq. kg⁻¹_2,3-BDO, allocated to the production of A-molasses feedstock and bagasse. The emissions from the sugar mill were identical for all scenarios in the FFD study with the remainder being from 2,3-BDO production in the biorefinery.

Overall, emissions from sugarcane cultivation and harvesting accounted for 53.6% of the total emissions from the sugar mill section as shown in Fig. 6. This was due to the usage of fertilizers and pesticides made from high emission factor chemicals such as urea (3.31 kgCO₂eq. kg⁻¹) and triazine (7.98 kgCO₂eq. kg⁻¹), respectively. Other notable emissions from sugar mill section were traced from lime usage for sugarcane juice clarification, bagasse combustion in the CHP plant and fossil fuels used for feedstock transport and blending. Emissions from bagasse usage were designated as biogenic and were allocated a reduced overall GHG effect thus accounting for only 17.4% of total sugar mill emissions.

The impact of bioprocess performance on total GHG emissions (sugar mill \(+\) biorefinery) is depicted in Fig. 6b. The overall GHG emissions per kilogram of 2,3-BDO were generally decreased by improvements to titre, to a more significant extent than improvements in product yield and volumetric productivity; an observation that was consistent with the findings from a similar study on biobased acrylic acid (Bhagwat et al. 2021). For instance, regression coefficients analysis showed that a 10% increase in product yield (at fixed titre) reduced the GHG emissions by 2.1% compared to a 4.9% reduction in GHG emissions for a 10% improvement in titre (at fixed product yield). The usage of process water and subsequent production of wastewater reduced notably at high titres lowering the GHG emissions associated with WWT treatment. In addition, the reduced overall heating demand for the biorefinery at high titres lowered the bagasse consumption rate and associated CO₂ emissions in the DOWTHERM section, leading to a low overall GHG emission (data not shown).

The total emissions per unit product reduced with an increase in product yield due to increased 2,3-BDO production rate from the fixed/constant supply of A-molasses, and decreased formation of acidic coproducts, particularly acetic acid, reducing the requirements for lime in the acid digester. However, the titre still overshadowed the product yield with respect to GHG emission benefits due to its much superior influence on overall heating requirement compared to the other energy demand which were closely shared between the two metrics (see- Section: 2.1). Because the volumetric productivity had a negligible direct effect on mass and energy flows per unit of 2,3-BDO (see—“Overview of methodology” section), it equally showed no impact on GHG emissions overall. In real situations, however, changes in the fermentation time may affect the amount of energy lost from heated process equipment such as bioreactors, which may affect GHG emissions.

A further examination of Fig. 6 (showed that the contour lines for the GHG emissions became steeper above the 100 g_2,3-BDO  L⁻¹ region indicated by the dashed line which indicates a region of improved GHG benefits from increments in titre than product yield. In this region, regression analysis showed that a 10% improvement in titre alone reduced the GHG by 8.3% which was four times more than those from a 10% increment in product yield (2.8%). Enhancement of product titre is thus a priority for greener production of 2,3-BDO in comparison with product yield and volumetric productivity. Bioprocess strategies to improve volumetric productivity at the expense of titre and product yield are therefore not preferred, especially for processes below the dashed line (see—Fig. 6b).

Finally, the potential lowest GHG emissions projected from this study was 0.77 kg-CO₂eq. kg⁻¹_2,3-BDO achievable at product yield and titre values of 0.495 g_2,3-BDO  g_sugar and 150 g_2,3-BDO  L⁻¹. These emissions are roughly sevenfold lower than the GHG emissions reported for similar products like fossil-based 1,4-BDO (Rehman et al. 2020). To assess the current state of technology for biobased 2,3-BDO, actual bioprocess parameters from key studies on microbial 2,3-BDO were used as inputs to the models developed in this study and GHG emissions calculated as before (see—Table 5). Results showed that bio-based 2,3-BDO promises significant environmental merits with GHG emissions ranging between 0.85 and 1.10 kg-CO₂eq. kg⁻¹_2,3-BDO  There are opportunities to further reduce the GHG emissions though final titre enhancements by engineering high yielding organisms capable of attaining titres above 150 g_2,3-BDO  L⁻¹ and optimizing feeding strategies by using fed-batch cultures to improve both titre and product yield. In addition, there is need to develop more efficient 2,3-BDO extraction and purification methods which eliminates the use of high emission factor chemicals such as lime and oleyl alcohol.