A Framework for Optimizing Agricultural Waste Biorefinery via Multi-Objective Modeling

Thesis Abstract

The rapid proliferation of agricultural residue streams and the rising demand for sustainable valorisation have intensified the need for integrated biorefinery frameworks that simultaneously optimize economic, environmental, and social objectives. Agricultural waste streams from cereal crops, fruit pomace, and oilseed pressings present a complex mix of sugars, lignocellulosic fibers, lipids, and phenolics, whose valorisation potential is often constrained by competing end-use pathways, resource constraints, and market dynamics. This study aims to develop and validate a robust multi-objective optimization framework that guides the design and operation of agro-waste biorefineries to maximize net present value, reduce life-cycle greenhouse gas emissions, and enhance rural employment. The specific objectives are (i) to formulate a multi-objective mixed-integer linear programming (MILP) model that integrates sequential pretreatment, conversion, and product-pipeline decisions under techno-economic and environmental constraints; (ii) to incorporate risk-adjusted performance via a stochastic programming extension capturing feedstock variability and market price fluctuations; (iii) to embed a sustainability assessment module based on life-cycle assessment (LCA) and multi-criteria decision analysis (MCDA) to support stakeholder trade-offs; (iv) to validate the framework using a representative district-scale dataset comprising 12 crop residue streams and 4 biorefinery configurations; and (v) to develop a software prototype with user-friendly interfaces for scenario analysis and policy guidance. The research adopts a quantitative, design-optimization paradigm under a pragmatic positivist stance, drawing on the theoretical underpinnings of multi-objective optimization theory, process systems engineering, and sustainability science. The population comprises agricultural processing facilities within a temperate agrarian region, with sample inputs including feedstock production data, collection costs, processing yields, energy balances, capital expenditures, and market prices collected from 2019–2024. Data collection instruments include official agronomic and supply-chain databases, facility annual reports, and structured expert elicitation with 15 professionals spanning farmers, processors, and policy makers. Validity and reliability are ensured through triangulation of market data, cross-validation with 3 case-study biorefineries, and test–retest reliability checks on survey instruments. The MILP model is implemented in GAMS and tested in MATLAB, with sensitivity analyses conducted via Latin Hypercube Sampling and scenario analysis across 25 representative futures. The optimization framework integrates a production-planning module, a capital-investment module, and an environmental module, with objective functions for net present value, life-cycle greenhouse gas emissions, and job-years created. The methodology includes a stochastic extension to address feedstock variability and price uncertainty, using scenario-tree representations and chance-constrained constraints. Hypothesis testing focuses on whether the multi-objective model significantly improves economic performance while reducing environmental impacts relative to baseline single-objective configurations, evaluated through paired t-tests and non-parametric equivalents when necessary. Expected findings indicate that integrated pretreatment choices coupled with bioproduct diversification (e.g., fermentative bioethanol, lipid-based compounds, and lignin-derived materials) yield superior trade-offs, with a 12–22% increase in NPV and a 8–15% reduction in life-cycle emissions under mid-range feedstock variability. The framework is anticipated to reveal critical levers such as feedstock blending ratios, centralized vs. decentralized processing, and feedstock storage strategies that minimize risk without compromising output quality. The study contributes to knowledge by providing a transparent, transferable framework that links process-engineering decisions to sustainability outcomes in agricultural waste valorisation, offering a reproducible methodological blueprint for regional policymakers and industry stakeholders. It advances the literature on agro-waste biorefineries by integrating multi-objective optimization with LCA and MCDA, addressing a recognized gap in operationalizing sustainability metrics within process design. The principal conclusion is that a coupled MILP-stochastic framework enables explicit exploration of trade-offs and supports evidence-based decision-making for resilient, value-enhancing biorefineries. Recommendations include adopting region-specific feedstock profiling, investing in modular processing steps to accommodate variability, and developing a policy package that incentives co-production of high-value bio-based products while maintaining competitive energy and waste-management costs. Further research is suggested to extend the framework to dynamic, real-time control under stochastic disturbances and to incorporate social acceptance indicators into the MCDA layer.