Hydrogen trilism: a context-sensitive framework for integrating renewable energy, socio-economic realities, and policies for clean cooking transitions

Cooking is fundamental to human well-being, ensuring food safety and nutrition. Yet, despite global efforts to expand clean energy access, over 2.4 billion people in the Global South still rely on inefficient, polluting cooking methods. This lack of access drives deforestation, indoor pollution, economic stagnation, and severe health risks ' contributing to 3.2 million premature deaths annually from household air pollution. Addressing these challenges requires innovative and scalable solutions that are feasible for end users in low- and middle-income country (LMIC) settings. This thesis investigates hydrogen as a contested clean cooking energy pathway while explicitly foregrounding barriers that may limit plausibility, including end user experience, emissions trade-offs, high pressure storage and cylinder logistics, and affordability relative to incumbent fuels.

Despite its potential, the feasibility of hydrogen remains highly debated due to conflicting assessments of costs, scalability, and policy effectiveness. At household scale, the feasibility debate also centres on end use performance and user experience, because the available evidence base is thinner for domestic cookers than for large scale and industrial hydrogen burners. In addition, practical delivery for cooking implies high pressure storage and cylinder filling requirements, which introduces challenges for transport and handling in LMIC contexts. These consumer side and logistics constraints matter as much as system level cost projections when judging plausibility. Existing energy system optimisation models oversimplify geographical and infrastructure contexts, neglect financial constraints, and fail to capture the complexities of energy poverty in the Global South. These models often ignore the non-linear interactions between hydrogen production, electricity generation, and policy effectiveness, leading to evaluations that lack contextual validity.

Recognising these limitations, this thesis proposes a shift toward context-sensitive modelling that better reflects the realities of the Global South. To achieve this, it introduces hydrogen trilism ' a framework capturing the interdependencies among energy poverty, hydrogen systems, electricity infrastructure, and policy influence. Each component is interconnected: hydrogen production depends on electricity, which is shaped by existing infrastructure and policy conditions. Meanwhile, energy poverty and consumption patterns feed back into this system, influencing demand, production, and policy responses. These dynamics highlight the need for an integrated approach when evaluating hydrogen feasibility.

The triadic framework is structured into three phases (Chapters 4'7). Phase 1 (Chapter 4) applies a data-driven approach to understanding clean cooking access in Africa. A CatBoost machine learning model, optimised using Optuna and explained with SHapley Additive exPlanations (SHAP), was developed to predict the key drivers of clean cooking adoption. Findings from this phase challenge the conventional assumption that electricity access directly leads to clean cooking adoption. Instead, the results reveal that financial resilience (the ability of households to afford electricity or clean cooking fuels) is a far stronger determinant. These insights underscore the need for a holistic approach to addressing energy poverty, prioritising affordability and economic resilience. Governments and development agencies can drive a much-needed paradigm shift in energy poverty mitigation by integrating these findings into policy frameworks.

Phase 2 (Chapters 5 & 6) focuses on modelling, validation, and multi-objective particle swarm optimisation of a renewable hydrogen system, incorporating real-world technical constraints such as dynamic electrolysers and fuel cells. To integrate ML-driven insights from Phase 1 into system optimisation, a new socio-economic metric, Levelised Value Addition (LVA), was introduced to assess the broader socio-economic benefits of energy systems alongside technical and cost-based metrics. Additionally, a top-down load profiling approach was developed and benchmarked against the Human Development Index to determine the energy consumption levels required for self-sustaining clean cooking ecosystems, as identified in the ML-based clean cooking pathways. To further enhance decision-making, the Data-Driven Multi-Objective Strategy for Sorting Solutions (DDMOSSS) was introduced. This novel framework maps SHAP feature values onto objective functions, allowing for an evidenced-based ranking of Pareto-optimal solutions.

The findings revealed that while DDMOSSS ranks solutions similarly to Technique for Order of Preference by Similarity to Ideal Solution under large-scale configurations, DDMOSSS is better suited for small-scale systems where socio-economic trade-offs are more pronounced. The financial analysis revealed that non-islanded large-scale hydrogen systems using imported electricity can achieve competitive hydrogen selling prices of $7/kg. This finding contributes to the debate on islanded versus non-islanded and non-trade islanded configurations, showing that non-islanded solar PV systems outperform islanded and non-trade islanded configurations. Additionally, while Production Tax Credits (PTCs) significantly reduce the levelized cost of hydrogen and net present cost by over 113% from baseline values, the LVA increases by only 24% because socio-economic benefits are constrained by physical and technological limitations of a system rather than purely financial incentives. In short, PTCs primarily shift the producer side net cost metrics into a net positive value creation region (cash flows and profitability) rather than directly benefiting consumers. Furthermore, while Investment Tax Credits provide upfront cost savings, Carbon Credits offer sustained financial benefits by aligning revenue streams with hydrogen consumption. The results also showed that achieving hydrogen market competitiveness requires scaling demand, incorporating policy incentives, and driving technological advancements. For instance, hydrogen must reach $1/kg to compete with charcoal in Zambia or $5/kg to compete with LNG.

Phase 3 (Chapter 7) focuses on expert stakeholder engagement, a critical element in the hydrogen trilism framework, where the interdependence of energy poverty, hydrogen systems, and electricity infrastructure shapes the feasibility of hydrogen as a clean cooking solution. Stakeholders (both human and institutional) play a pivotal role in influencing these interactions, as decisions in one area can create cascading effects across the entire system. To account for these complexities, this phase integrates systematic stakeholder-driven impact assessments, pre-emptive and mitigative policy analysis, and system design optimisation.

The findings confirm that hydrogen cooking faces substantial barriers in LMIC contents and that any role it may play is likely to be limited to specific settings and enabling conditions. Several unwanted impacts and key barriers must be addressed, including disruptions to traditional fuel practices and livelihoods, cultural resistance to fuel-switching, high hydrogen fuel costs, lack of technical expertise and infrastructure constraints. To overcome these challenges, pre-emptive and mitigative policies are essential. These include alternative livelihood programmes, incentives for cultural acceptance and fuel-switching, technical capacity building, and infrastructure development. Furthermore, integrating stakeholder preferences into system design is crucial. While DDMOSSS performs well in small-scale scenarios, larger systems require a more locally tailored approach that explicitly incorporates regional socio-economic preferences.

The findings provide compelling evidence for the hydrogen trilism thesis: the feasibility of hydrogen cannot be meaningfully assessed without considering the interactions between energy poverty, hydrogen systems, and electricity infrastructure. This thesis demonstrates that hydrogen adoption is not purely a cost-driven challenge. It requires a holistic, systemic approach that integrates context-sensitive modelling, socio-economic factors, and scale-dependent technology configurations. These findings offer practical insights for policymakers, investors, and development agencies, reinforcing that overcoming energy poverty and achieving universal clean cooking access demands solutions that extend beyond traditional cost-based frameworks.