Optimizing the IFE design for a fusion pilot plant

The 2021 report of the National Academies discusses a fusion pilot plant producing 50 MW of electricity for 3 hours. To meet such power requirement using the inertial confinement fusion approach, we propose to investigate promising designs at the hypothetical 5 MJ laser facility using radiation-hydrodynamic simulations. Our initial goal is to find a design that not only produces high yield, but also does so robustly, that is, with little sensitivity to the uncontrolled variability in the laser delivery and target quality. We leverage an iterative optimization workflow developed by the LLNL’s ICeCap team, which involves running integrated laser-hohlraum-capsule simulations in batches, building a machine learning surrogate model of these simulations, and in-line proposing new samples of the design inputs automatically. Some of these samples refine promising designs, while others reduce the model uncertainty in poorly explored regions, ultimately finding the optimal design with far fewer expensive simulations compared to random sampling of the design space. We will discuss the physics constraints applied during the optimization as well as tradeoffs between different parameters. Finally, we demonstrate that the optimal design exceeds the 50 MJ of output energy (target gain of 10), which, assuming a 10 Hz rate and 10% efficiency, would produce the required 50 MW of electricity. Future work will include defining laser driver margin for variability to ensure the required gain for power generation.