Kibaek Kim, an assistant computational mathematician in Argonne National Laboratory’s Mathematics and Computer Science Division, has received an Early Career Research award from the U.S. Department of Energy’s (DOE) Office of Science.

The Early Career award identifies and supports exceptional, promising young researchers. Kim will receive $2.5 million over five years for his research on data-driven optimization under uncertainty.

The key challenges in the design and control of cyber-physics systems are threefold: the complex physics, network topology and logics governing the systems; the need for computationally expensive simulations of operations for long periods of time; and uncertainties such as extreme weather and component failure affecting design and control decisions.

“Data-driven optimization under uncertainty can address these challenges by developing models and algorithms in distributionally robust multistage stochastic optimization,” Kim said.

In particular, Kim and his team are exploring a new approach that makes multistage decisions under uncertainty from an ambiguity set of distributions, in which uncertain events occur sequentially over multiple periods of time. He noted that existing solutions are limited to serial computing algorithms or to two-stage problems and that they make limited use of high-performance computing (HPC) for optimization.

“Our aim is to devise parallel algorithms and scalable solvers that are statistically robust to changes of the distribution and that can exploit advanced HPC architectures equipped with accelerators,” Kim said. Essential to the approach is the use of massive amounts of data to achieve greater out-of-sample performance and to avoid overfitting effects.

Kim plans to focus on two design and control applications: resilient smart distribution systems and distributed learning on edge computing networks.