Topology optimization for energy problems

The optimal design of energy systems is a challenge due to the large design space and the complexity of the tightly-coupled multi-physics phenomena involved. Standard design methods consider a reduced design space, which heavily constrains the final geometry, suppressing the emergence of design trends. On the other hand, advanced design methods are often applied to academic examples with reduced physics complexity that seldom provide guidelines for real-world applications. This dissertation offers a systematic framework for the optimal design of energy systems by coupling detailed physical analysis and topology optimization. Contributions entail both method-related and application-oriented innovations. The method-related advances stem from the modification of topology optimization approaches in order to make practical improvements to selected energy systems. We develop optimization models that respond to realistic design needs, analysis models that consider full physics complexity and design models that allow dramatic design changes, avoiding convergence to unsatisfactory local minima and retaining analysis stability. The application-oriented advances comprise the identification of novel optimized geometries that largely outperform industrial solutions. A thorough analysis of these configurations gives insights into the relationship between design and physics, revealing unexplored design trends and suggesting useful guidelines for practitioners. Three different problems along the energy chain are tackled. The first one concerns thermal storage with latent heat units. The topology of mono-scale and multi-scale conducting structures is optimized using both density-based and level-set descriptions. The system response is predicted through a transient conjugate heat transfer model that accounts for phase change and natural convection. The optimization results yield a large acceleration of charge and discharge dynamics through three-dimensional geometries, specific convective features and optimized assemblies of periodic cellular materials. The second problem regards energy distribution with district heating networks. A fully deterministic robust design model and an adjoint-based control model are proposed, both coupled to a thermal and fluid-dynamic analysis framework constructed using a graph representation of the network. The numerical results demonstrate an increased resilience of the infrastructure thanks to particular connectivity layouts and its rapidity in handling mechanical failures. Finally, energy conversion with proton exchange membrane fuel cells is considered. An analysis model is developed that considers fluid flow, chemical species transport and electrochemistry and accounts for geometry modifications through a density-based description. The optimization results consist of intricate flow field layouts that promote both the efficiency and durability of the cell.