Subventions et des contributions :

Subvention ou bourse octroyée s'appliquant à plus d'un exercice financier. (2017-2018 à 2022-2023)

The unique context of energy systems requires efficient, adaptable design tools able to quickly explore new areas and emerging problems. Two exciting new developments in machine learning, hyper-heuristics (optimizing the optimizer) and fitting of meta-models using statistical emulators, can be combined in a modular fashion to provide such tools.
Preventing disastrous levels of climate change, ensuring energy security and achieving a sustainable future all require novel energy systems. These will be less centralised and ‘top-down’ since local balancing of demand and supply is critical temporally and spatially. Analysis of these systems must span from buildings (which are now active players in energy markets) to district, city and national infrastructure.
This research program aims to investigate the holistic, integrated modelling, design and optimization of diverse energy systems. It will leverage the unique benefits of a modular modelling environment (mod2e), encompassing simulation and analysis, optimization, hyper-heuristics and statistical meta-models amongst others.
Existing simulation tools address single design options rather than extensive design-space exploration. There has been considerable success in applying computational optimization methods to find good solutions across a broad design-space, but further progress requires a different approach in which optimization and modelling are coupled more closely.
The underlying methodology is the modularisation of simulation, optimization and meta-modelling elements to form a modular modelling environment (mod2e). New and existing techniques will be combined more effectively, then applied to diverse energy systems problems with academic and commercial partners.
The modularisation of optimization will allow tuning of optimizers for particular sub-problems using hyper-heuristics, which extends the field of meta-heuristics to ‘optimizing the optimizer’. Statistical meta-models allow time-consuming simulations to be replaced by fast approximations which give adequate accuracy during the early stages of optimization.
Models can be easily reused and reconfigured to address wide-ranging design problems to meet new research challenges. This flexibility will enable an adaptive process rather than the solution of a static problem. Extensions will cover semi-autonomous module configuration and expansion to additional domains such as city planning and electric vehicle deployment.
The impact of the research program will be to deliver more useful, more powerful, more holistic simulations and optimizations by embracing modularity. Faster, more detailed exploration of complex design-spaces will allow broader questions to be explored in greater depth. This will enable significant improvements in how energy systems for buildings, districts and cities can be designed and operated to meet future challenges.