Rodgers, Sarah
(2024)
Techno-economic analysis and method development applied to an aerobic gas fermentation and supercritical water gasification process.
PhD thesis, University of Nottingham.
Abstract
The chemical and fuels industry's reliance on fossil-based feedstocks necessitates a shift to low-emission renewables like agricultural residues, municipal solid wastes, and industrial off-gases. Gas fermentation employs microbes to convert gaseous carbon-rich streams into renewable chemicals. The current state of commercial gas fermentation relies on anaerobic bacteria. However, aerobic bacteria offer the potential to target a broader range of products. Despite this, an inherent disadvantage of aerobic gas fermentation is its poor thermodynamic efficiency. Integrating aerobic gas fermentation with Supercritical Water Gasification (SCWG) addresses this challenge, creating a promising biochemical production platform. The integration utilises the low-temperature fermentation heat to pre-heat biomass for SCWG and reclaims the energy from depressurising the SCWG effluent using a turbo-expander. In addition, a key benefit of SCWG is its ability to exploit wet, low-value wastes. Such feedstocks are abundantly available and have limited resource competition as they are uneconomical to exploit via conventional gasification. Despite these benefits, the economic feasibility needs verifying which includes the selection of optimal feedstocks, feasible production capacities, and geographic considerations to identify promising biorefinery scenarios.
Essential for assessing emerging technologies, Techno-Economic Analyses (TEAs) were conducted to rigorously model and assess two case studies for the proposed integrated technology. The first considered commodity chemicals as direct products from fermentation (co-produced isopropanol and acetone), achieving a cumulative Net Present Value (NPV) of $42 million. Results compared favourably to anaerobic fermentation, with a minimum fuel selling price of $2.87/GGE. The second case study considered hybrid processing, integrating bio- and chemo-catalytic upgrading to produce 1,3-butadiene. This process showed profitability, achieving an MSP of $1367/tn, a $2.8M NPV, and a 19% probability of positive NPV. Part of the success of these two case studies was due the use of low-cost black liquor as the feedstock. SCWG allows for the successful exploitation of this wet feedstock. As such, a final study was undertaken to identify promising biorefinery scenarios for hydrogen production via SCWG, considering different feedstock-capacity-location combinations. The Levelised Cost of Hydrogen (LCOH) ranged from 3.81 to 18.72 $/kgH2 across the considered scenarios. At capacities >50 m3/h, the LCOH’s (2.76–4.21 $/kgH2 for China, 3.41–5.07 $/kgH2 for Brazil, 4.31–6.62 $/kgH2 for the UK) were competitive with MW-scale electrolysis costs (3.10–6.70 $/kgH2 for China, 3.70–5.90 $/kgH2 for Brazil, and 4.81–6.31 $/kgH2 for the UK). The range of results highlights the significance of feedstock-capacity-location considerations during technology evaluation.
In evaluating the economic feasibility of bio-derived chemicals and fuels, it's crucial to conduct Life Cycle Assessment (LCA) to quantify environmental impact. This facilitates a comparison of the trade-offs between a process’ economic and environmental performance. For both commodity chemical case studies net negative emissions were achieved due to biogenic carbon sequestration. Isopropanol and acetone exhibited GHG emissions of -2.10 and -2.21 kgCO2eq/kg compared to conventional production of 2.07 and 2.43 kgCO2eq/kg. For 1,3-butadiene production emissions were -3.23 kgCO2eq/kg, contrasting with the conventional 1.2 kgCO2eq/kg. Hydrogen production from the final case study also demonstrated low process emissions, averaging 0.46 kgCO2eq/GJH2 (China and Brazil), and 0.37 kgCO2eq/GJH2 (UK), compared to 8 kgCO2eq/GJH2 for steam methane reforming with carbon capture and storage (excluding natural gas leakage). These favourable emissions across all studies highlight the benefits of exploiting low-value, low-emission feedstocks.
As part of a TEA product prices for 20-25 years into the future are required to assess potential profitability. A Machine Learning (ML) method for projecting future commodity prices was developed to allow for unbiased price selection procedures to input into TEAs. Initially, a Radial Basis Function Neural Network (RBFNN) was trained using 10 historic prices, optimising weights and centre points. The model was run recursively, with predicted prices becoming inputs. Stochastic uncertainty was incorporated using a ±30/20% uniform distribution from the projected price. The method was later refined using 100 LSTM models, leveraging historic commodity data (2009-2021) and Energy Information Administration's (EIA) Brent crude oil price projection. Training and validation sets were based on a 30% historic data and 70% projection horizon ratio, ensuring optimal hyperparameter selection. Probabilistic projections provided nominal, range, and probability distributions to input into the economic, sensitivity, and uncertainty analysis. The resulting price distributions showed variability between commodities, emphasising the need for tailored TEA uncertainty considerations instead of relying on arbitrary percentages. Compared to the initial RBFNN method, the refined approach was found to alter the NPV distributions' 70% window from $35-$95M to $45-$80M (isopropanol and acetone) and from -$45M-$65M to -$35M-$80M (1,3-butadiene), highlighting the importance of price selection procedures on TEA outcomes.
Conducting TEAs is time consuming and requires expert knowledge, hindering widespread application. To facilitate quick biorefinery scenario evaluations a ML method was developed. This was created for the TEA of hydrogen production via SCWG. An ML surrogate model was developed to predict the LCOH based on different feedstock-capacity-location combinations. The training data included 40 biomass compositions, five processing capacities (ranging from 10 to 200 m3/hr), and three geographical locations (China, Brazil, UK). Three ML algorithms were compared for the ML surrogate model: Random Forests, Support Vector Regression, and an ensemble of Artificial Neural Networks (ANNs). The ANN ensemble was the most accurate during cross-validation and achieved an accuracy of Mean Absolute Percentage Error: <4.6%, Route Mean Squared Error: <0.39, and R2: >0.99 on the test set. The final model was published for users to evaluate their own feedstocks. Overall, the model enables the identification of promising biorefinery scenarios for valorisation to maximise the economic potential of the technology.
There are two key contributing areas of this thesis, firstly, the rigorous techno-economic and environmental assessment of the technology and secondly, the development of TEA methods using ML to aid these evaluations. The techno-economic and environmental assessment demonstrates the economic and environmental viability of the proposed technology platform compared to both conventional and alternative renewable production routes. The development of TEA methods used ML to create an unbiased methodology to select product price and price distributions in TEAs and to produce a TEA surrogate model for early-stage screening of feedstock scenarios for SCWG. The methods developed demonstrate the potential of ML to enhance TEA practices.
| Item Type: |
Thesis (University of Nottingham only)
(PhD)
|
| Supervisors: |
McKechnie, Jon
Lester, Edward |
| Keywords: |
Techno-economic analysis, life cycle assessment, renewable chemicals, biomass, process modelling, machine learning |
| Subjects: |
T Technology > TP Chemical technology |
| Faculties/Schools: |
UK Campuses > Faculty of Engineering > Department of Chemical and Environmental Engineering |
| Item ID: |
77516 |
| Depositing User: |
Rodgers, Sarah
|
| Date Deposited: |
18 Jul 2024 04:40 |
| Last Modified: |
18 Jul 2024 04:40 |
| URI: |
https://eprints.nottingham.ac.uk/id/eprint/77516 |
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
