*TotalEnergies is a registered trademark of Group TotalEnergies and it is reproduced with its owner's authorization.
Area: Energy
Problem class: MINLP
Technologies: GAMS
TotalEnergies is a global integrated energy company that produces and markets energies: oil and biofuels, natural gas, biogas and low-carbon hydrogen, renewables and electricity. Its more than 100,000 employees are committed to provide as many people as possible with energy that is more reliable, more affordable and more sustainable. Active in about 120 countries, TotalEnergies places sustainability at the heart of its strategy, its projects and its operations.
As part of its ambition of carbon neutrality by 2050, together with society, TotalEnergies develops Carbon Capture Storage (CCS) solutions –capturing CO2 emissions from industrial sites, transporting them, and permanently storing them beneath the seabed
To make this work technically and economically feasible, TotalEnergies developed a sophisticated GAMS optimization model to manage the complex operations of CO2 logistics: buffering in tanks, transport, and injection into subsurface reservoirs. This large-scale MINLP model was built upon the work and in collaboration with Professor Grossmann and several students from Carnegie Mellon University. However, taking the academic model and turning it into a practical industrial model presented significant challenges, particularly due its long-term planning horizon and detailed daily operational decisions.
GAMS Consulting addressed the model’s bottlenecks and made several breakthroughs. Through targeted improvements GAMS helped TotalEnergies achieve a performance gain that took the model from running a field development strategy life of 1-2 years in about an hour, to 20 years in a few minutes. Additionally, we developed a user-friendly Python interface that technical staff without expert domain knowledge can easily use and interact with.
The CCS optimization model developed by TotalEnergies was designed to operate through an important part of the carbon value chain –from CO2 arrival at port terminals to final injection in off-shore reservoirs. However, what made the model so powerful also made it impractical to use. Capturing the operational nuances of well pressure behavior, injection dynamics, and tank buffering across multiple time scales meant incorporating highly non-linear constraints and a large number of binary decisions. The result was a model with tens of thousands of variables and constraints that grew exponentially with each additional month of planning.
This complexity was exacerbated by the need to make decisions at daily intervals, while planning over a horizon of years or decades. Basically, the long-term vision collided with short-term computational feasibility: the model took too long to solve, especially when simulating realistic scenarios.
These delays made carrying out studies time-consuming and cumbersome. It was simply inefficient to run tests that took over an hour each. Additionally, the highly technical structure of the model made it difficult for non-specialists to interact with, which limited the accessibility and practical value for the engineers who needed it most.
Our team approached this problem with a strategy that blended technical expertise with a deep domain understanding and developed a solution in multiple steps.
As a first step, we conducted a thorough audit of the existing model to identify computational and structural bottlenecks. Then, the team refactored the model’s code using modern GAMS features to improve maintainability and streamline logic; constraints and time structures were reformulated by leveraging problem-specific knowledge, reducing the overall model’s complexity; and decomposition algorithms used to solve the model were improved, resulting in a significantly more efficient optimization process.
These efforts culminated in a dramatic performance boost. For a 25-year horizon model, the runtime went from hours to a few minutes. This improvement in runtime was most noticeable during the solve phase, which made it feasible to run detailed, realistic scenarios in minutes. Plus, this meant that the model could be improved even further in complexity –for example, allowing for uncertainties, which just wasn’t a possibility before.
Recognizing that the model would be used primarily by reservoir engineers –professionals more focused on operational insight than programming– our team redesigned the user interaction experience. We developed a Python-based interface that allows users to configure model runs, manage input and output data, and view results in a structured and intuitive way. Then, in order to bridge the gap between advanced optimization and user-friendly operation, this interface was paired with an Excel-compatible frontend. This allowed engineers to interact with the model using familiar tools, removing the need to understand GAMS syntax or Python code, and thus enabling a more seamless workflow between the engineers and the optimization platform.
To support long-term sustainability and encourage adoption, we also prioritized training and knowledge transfer. The consulting team delivered detailed documentation covering the model’s architecture, usage of advanced GAMS features, and run configuration. We also produced a simplified user manual with examples, operational guidelines, and troubleshooting tips, ensuring that TotalEnergies’ employees could independently maintain and adapt the model for future needs.
This partnership showcases the real-world impact of combining advanced optimization with expert consulting. TotalEnergies was faced with a highly complex CCS model that was difficult to maintain and underperformed in terms of computational time, and through targeted model refactoring and algorithmic improvements, the GAMS team reduced runtimes significantly –enabling realistic scenario planning and faster decision-making. This speedup also allows TotalEnergies to add even more complexity to their model, relaxing the existing assumptions and getting the model as close to reality as possible.
Equally important was the redesign of the user interface, which made the model accessible to engineers without deep programming expertise. By integrating a Python backend with an Excel-compatible frontend, GAMS ensured that the tool could be used directly by those closest to the operational decisions.
The result was the transformation of a techincally powerful but cumbersome model, into a streamlined, strategic tool. Our team not only accelerated the computations, but also gave the engineers at TotalEnergies a solution fast enough to use day-to-day, changing how they approach optimization to support their CCS operations.
Today, the model works as a robust, user-friendly decision-support system that plays a key role in advancing TotalEnergies’ net-zero strategy. This collaboration is a clear example of how the right expertise can turn complexity into capability, and vision into action.
Ambrish Abhijnan, Kathan Desai, Jiaqi Wang, Alejandro Rodríguez-Martínez, Nouha Dkhili, Raymond Jellema, and Ignacio E. Grossmann
Industrial & Engineering Chemistry Research 2024 63 (27), 12053-12063
DOI: 10.1021/acs.iecr.4c00390