ABSTRACT

Recently the world is facing many unprecedented disruptions, such as extreme natural disasters and pandemics. In many realistic situations, the uncertainty may not occur frequently but can bring significant changes to system parameters. For example, the wildfire in California has caused damage to the power systems. A conventional sequential decision problem, where uncertainty takes place in every time period with random magnitude, does not model this uncommon type of event with the best computational efficiency. We propose a new mathematical model, stochastic disruptions, of which both the timing and the magnitude are random. We present two concrete examples of stochastic disruptions applied in different engineering applications. In the first example we discuss an optimal power flow problem for a distribution electricity network, with generators and lines potentially failing due to disruptions. In the second example we minimize the project span with a limited budget for acceleration, where the disruption can significantly change the duration of some activities. We also discuss potential extensions to the modeling framework on how we can construct a data-driven distributionally robust model for the timing randomness and utilize rare event simulation.

BIOGRAPHY

Dr. Haoxiang Yang is Postdoctoral Research Associate of the Center for Nonlinear Studies at Los Alamos National Laboratory. He graduated with a PhD degree from the Department of Industrial Engineering and Management Sciences at Northwestern University. His research focuses on the theory of optimization under uncertainty, specifically in stochastic programming and robust optimization, integer programming and nonlinear programming, with its applications in energy systems, supply chain and logistics, disaster relief and sports analytics.