University of Glasgow
Environmental and techno-economic analysis of biomass and waste gasification facilitated by machine learning
Abstract
dc:description.abstractBioenergy is essential for limiting global warming to 1.5°C, the goal set out by the Intergovernmental Panel on Climate Change (IPCC) in 2015. Gasification is one type of technology that can deliver bioenergy by converting biomass into fuels and chemicals via thermal conversion. When compared to other conversion technologies, gasification stands out due to its potential to efficiently turn biomass into syngas and biochar, which have a plethora of applications. Therefore, this work sets out to improve our knowledge of biomass and waste gasification (and its process performance and efficiency), thereby encouraging its widespread commercialisation and industrial-scale implementation. This was done by developing a comprehensive modelling framework to holistically assess the environmental and techno-economic performance of a biomass or waste gasification scheme, as particularly the trade-offs between environmental and techno-economic factors remain poorly understood. As a first step, a comprehensive data set of 312 samples was collected from the existing literature, to develop machine learning (ML) models to predict ten key biomass and waste gasification outputs (e.g. syngas yield and composition), which are vital for environmental and economic analysis. Gradient boosting regression was found to be highly accurate with a coefficient of determination of (R2 ) of 0.90 when averaged across all model outputs. The developed models also boast a better generalisation capability than existing models, meaning they can predict gasification performance over a wide range of input parameters, such as feedstocks, reactor types, and gasifying agents. Crucially, investors’ and policymakers’ trust in the models was ensured by incorporating interpretability methods. These methods highlighted the feedstock’s particle size and the gasifying agent as parameters of particular importance. The developed ML models were then used to establish an environmental and technoeconomic modelling framework, which utilises Monte Carlo methodology to account for sources of uncertainty within the framework. By implementing the framework through an easy-to-use graphical user interface, non-expert stakeholders can easily use the tool to rapidly assess the performance of a range of different gasification feedstocks and operational choices. Ultimately, the developed modelling framework was used to conduct a case study analysing three promising Scottish feedstocks (forestry residues, barley straw, and distillers’ malt draff). Through multi-objective optimisation and sensitivity analysis, the best feedstock and process conditions were identified. When considering carbon capture and storage (CCS), a maximum carbon avoidance of -2,948 kg CO2-eq/tonne of feedstock at a benefit-cost ratio (BCR) of 1.1 was found for the feedstock forestry residues. Other solutions considering CCS showed an improved economic performance of BCR = 1.41, despite a slightly diminished environmental performance of approximately -2,400 kg CO2-eq/tonne of feedstock. Optimal solutions not considering CCS still led to negative emissions of up to -950 kg CO2-eq/tonne and an improved economic performance (BCRs ranging from 1.72 to 1.85). The case study effectively illustrates how the developed modelling framework can be used to rapidly and easily compare a range of feedstocks and gasification system choices. It also lays a foundation for more expansive modelling frameworks in the future which could consider a range of other bioenergy options, aiding with decisions on how to best utilise our limited biomass resources.
Degree
thesis:*- Level dc:type.qualificationlevel
- PhD
- Grantor dc:publisher.institution
- University of Glasgow
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ascher, Simon
Subjects
dc:subject × 2Rights
- Language dc:language
- en
Identifiers
dc:identifier.*- DOI dc:identifier.doi
- https://doi.org/10.5525/gla.thesis.84564