The project pursues three primary research thrusts:

• embedding process-level models of key industrial technologies into national-scale energy system models;
• developing computational methods, graph-based, decomposition, GPU-accelerated optimization, and machine learning,  to enable high-resolution, coupled energy and industrial commodity network modeling;
• applying these methods to a case study of the U.S. chemicals and fuels sector to identify feasible, viable, and well-timed decarbonization strategies.

CMCC role
CMCC is a co-host institution and will develop high-resolution, optimization-based network models with advanced computational performance.
CMCC’s team will be mainly involved in Research Thrust 2 “Graph-Based Decomposition & Aggregation Methods,” leading task 2.2 “Graph-Based Aggregation for Faster Convergence” and task 2.4 “ML-Accelerated Benders Convergence,” and contributes to task 2.3 “Time-Series Aggregation Leveraging Autoencoders,” as well as Research Thrust 3: Modeling the U.S. Chemicals and Fuels Sector in Mid-Transition, contributing to task 3.3 “Analyzing Competing Policy Priorities”.

Activities
The main activities in the first phase of the project will focus on developing bottom-up process models for conventional and emerging fuel and chemical production technologies, combining first-principles engineering with data-driven approaches to characterize facility-level costs, material flows, and performance (Task 1.1). These models will then be distilled into compact, tractable representations through facility clustering, reduced-order surrogate models, and mixed-integer formulations, for integration into the open-source MacroEnergy.jl platform (Task 1.2). In parallel, the team will accelerate the underlying optimization methods by implementing GPU-based solvers using the MPAX library (Task 2.1), designing graph-based aggregation techniques for multi-commodity time-space networks (Task 2.2), applying autoencoders to improve time-series aggregation (Task 2.3), and training machine learning classifiers to speed up Benders decomposition convergence (Task 2.4). These technical and computational advances will be applied to characterize the existing U.S. chemicals and fuels production fleet through facility clustering and archetype modeling (Task 3.1), to build site-specific supply curves for alternative feedstocks such as biomass and waste plastics (Task 3.2), and to evaluate how industrial decarbonization intersects with competing policy goals like reshoring, trade, and circularity, guided by a Stakeholder and Technical Advisory Group (Task 3.3). Throughout the five years, the project aims to  organize and host an annual one-day global technical workshop alongside the Net-Zero X Global Initiative conference (Task 4.1).

Expected results
Three main outcomes are anticipated:
1. Improved process representation: computationally scalable, rigorous and well-documented methods, software implementations and data sets for industrial sector modeling in macro-energy systems.
2. Improved computational methods: By mid-point, advanced decomposition methods will enable unprecedented resolution in energy and industrial system models; by completion, our open-source tools support broader adoption of highly detailed, large-scale models.
3. Applied case study: comprehensive study of U.S. chemicals and fuels decarbonization pathways and competing drivers influencing the sector’s evolution.

Coordinating organization
Princeton University (Lead Institution), led by PI Prof. Jesse D. Jenkins.

Partners
Massachusetts Institute of Technology (MIT)
New York University (NYU)
University of Notre Dame (UND)
Euro-Mediterranean Centre on Climate Change Foundation (CMCC)