Context:

With decarbonation a must, oil shall let electricity and hydrogen become the two future energy carriers. If not stemming from hydrocarbons, hydrogen is intended to be generated by electrolysis, thanks to RES generated power surpluses. Hydrogen and electricity can be alternatives in end-uses or as intermediaries.

The emergence of such an integrated system will involve, integrate and/or transform all other present energy carriers (e.g. methane) and/or networks at local, regional and global level. Hydrogen is however actually only the keystone in an already ongoing sector coupling (electricity, gases, heat, industry, transportation etc.).

Sector coupling is expected to help increase the share of intermittent sources and the energy system overall efficiency. However, such coupling is highly complex: extended and interlinked time and geographical scales, systemic effects, multiplicity of actors and value chain stage, investment risks and capital intensity, regulation and technology uncertainty etc. How to plan for sector coupling giving such complexity?

Planning studies (e.g. expansion, adequacy) often rely on so-called Energy System Optimisation Models (ESOMs) which can represent energy (esp. power) systems at national and European scales, over long-term horizons (tens of years). They are based on mathematical programming formalisms, and rely on perfect market and perfect foresight assumptions. They ensure a certain fidelity with respect to reality by accounting for hourly energy balances and technical constraints. Their complexity implies challenging trade-offs between fidelity to and tractability (but also interpretability). This challenge has been and continues to be extensively studied in the literature. Collaborative studies are even more challenging when tools, knowledge and data are spread among the actors.

One way of tackling the abovementioned challenges is to “couple” modelling tools. Successful examples from the literature, e.g. for expansion planning, illustrate difficult and case specific implementations, especially when defining the feedback loop.

Objectives:

The MuESSLi (Multi-Energy System Smart Linking) project aims to bring model coupling to another level by building data-driven proxies of models before coupling them. “Smart-linking” will allow for solving bigger problems, while ensuring cooperation between actors with a certain level data privacy. The 5 PhD topics respectively focus on:

• Building sector-specific, MILP representable proxies of operational ESOMs to be integrated into the model of the main sector of interest; achieved by a machine learning + optimisation pipeline to reduce business errors metrics.
• Building sector-specific, MILP representable proxies of operational ESOMs to be integrated into the model of the main sector of interest; achieved by extracting information from decomposition methods (e.g. Benders, ADMM).
• Accelerating of the operational assessment of energy systems based on active learning and multi-fidelity modelling, with a specific application to the computation of the state-of-charge value of long-term storages.
• Modeling market imperfections with Stackelberg games (involving equilibrium problems with equilibrium constraints); and building proxies of the later to be integrated into ESOMs.
• Bi-directional coupling between multi-sector expansion and balancing models. The latter representing the activation of reserves to cope with forecast uncertainties; To better account for flexibility needs in ESOMs.

+ Beyond research works:

• Architecture a new fully open-source platform to model, optimise and better plan the future multi-energy system, relying on existing tools and models;
• Foster a living opensource project and community of users and developers to populate it and enrich it, taking advantage, when time comes, of new mathematical tools, increased computing ability, etc.

Project partners:

The project now gathers five PhDs in three universities (DTU, TU Delft, UP Comillas) and three industrial partners (RTE, TotalEnergies and NaTran).

Credits: pictures are from Jason Mavrommatis, Tommy Krombacher, Shaun Dakin from Unsplash