Energy system modelling for the decarbonisation of industrial process heat

Abstract

Industrial process heat is a major source of fossil fuel use and carbon emissions, yet it is often represented only coarsely in large-scale energy system models. This limits the ability of existing models to assess site-level fuel-switching opportunities, spatial resource allocation, and the wider electricity system implications of industrial decarbonisation. In particular, modelling industrial process heat requires approaches that can capture the diversity of industrial demand sites, the location of fuel and electricity supply options, and the infrastructure needed to connect them. This thesis develops a spatially explicit energy modelling framework to evaluate nationwide fuel-switching pathways for site-level industrial process heat up to approximately 500°C, corresponding to temperatures that can be supplied by high-pressure steam.

The framework combines P-graph superstructure generation with an automated translation tool that converts the structural representation into a Pyomo-based generalised disjunctive programming model. This enables large-scale structural optimisation using reusable modular components and multiple simulation modes. A modular architecture is developed to represent model components in a scalable and reusable manner. The framework is further supported by geospatial preprocessing, standardised input-processing tools, and output reporting modules, which help facilitate the integration of high-spatial-resolution data within a unified P-graph and algebraic modelling workflow.

The methodology is demonstrated through a New Zealand case study that compiles data with high spatial granularity on industrial process heat demand, electricity generation assets, biomass resources, grid infrastructure, and geospatial linkages between supply and demand nodes. The energy systems model is evaluated using three simulation modes: deterministic multiperiod optimisation, sequential deterministic optimisation, and Monte Carlo uncertainty analysis. Together, these modes allow the framework to evaluate systems-level least-cost decarbonisation pathways, grid reinforcement upgrade requirements, and the robustness of system configurations under uncertain input conditions. The deterministic multiperiod factory model captures temporal variation in demand and supply conditions. Building on the multiperiod model, the sequential framework represents staged investment and includes grid reinforcement assessment, while the Monte Carlo analysis evaluates uncertainty in key modelling parameters.

The results indicate that biomass resources play a dominant role in the system-level least-cost industrial process heat decarbonisation in the New Zealand case study. Electrification increases industrial electricity demand by approximately 3.5–9.2 TWh, depending on scenario assumptions, creating additional pressure on generation and grid infrastructure. Sequential analysis shows that grid reinforcement costs can be represented through an adjusted electricity price coefficient, while Monte Carlo analysis provides a distributional view of electrification uptake, biomass utilisation, and spatial outcomes under uncertainty. These findings show that both biomass and electrification should be viewed as complementary decarbonisation pathways in the industrial process heat space, with their relative roles shaped by local resource availability, electricity costs, and infrastructure constraints.

The central contribution of the thesis is a spatially explicit energy modelling framework that combines P-graph superstructure generation with a translation tool converting the structural representation into a Pyomo-based generalised disjunctive programming model. At the same time, the framework demonstrates how P-graph can be extended to support large-scale, spatially explicit, and uncertainty-aware energy system analysis.