Title:Human-in-the-loop MGA to generate energy system design options matching stakeholder needs

Abstract:Using cost-optimisation to support energy system design decisions hides from view many more feasible designs that, while economically comparable to the least-cost option, may have other practical advantages that stakeholders prefer. Increasingly, modellers look at Modelling to generate alternatives (MGA) as a way to go beyond few, optimal designs and provide stakeholders with broad portfolios of design options to appraise based on their preferences and knowledge of the problem. In practice, however, generating and exploring all the feasible system design options is not computationally possible. Modellers need to choose what design feature to generate diversity around, and yet, they cannot know what features may matter the most to heterogeneous real-world stakeholders. Stakeholders' knowledge is thus needed to guide the MGA search towards the most important trade-offs, but how to incorporate such knowledge in the MGA workflow is unclear. To address this gap, we propose a first-of-its-kind human-in-the-loop (HITL) approach that seamlessly and automatically integrates stakeholder preferences into an MGA workflow. In a nutshell, we elicit stakeholders' system design preferences via a first interaction between them and a tentative MGA design space. Hence, we decode those preferences so that it becomes possible to feed them back to the MGA algorithm to perform a guided search. This search produces an updated, human-trained design space with a higher share of designs that align with the elicited preferences. In a controlled experiment for the case study of the Portuguese energy system, we demonstrate that our HITL-MGA approach may facilitate the identification of a consensus solution that balances many conflicting stakeholder preferences. The approach shows promise for generalisation and real-world application to accelerate technically and socially feasible energy transition decisions.