Overview

In September 2009, the Department of Energy awarded a three year grant to George Mason University, jointly with Noblis titled “New Approaches for Rare-Event Simulation and Decision Making”.  The objective of the research is to efficiently determine via simulation the best choice among a number of alternatives, where the simulation of each alternative requires the evaluation of a rare-event probability. We will apply our new rare-event simulation and decision making approach to two critically important problems – specifically, in the areas of network security and electric power grid control.

Simulation is a powerful tool that can be used to analyze a wide variety of systems. In principle, given an accurate model and ample computer time, simulation can provide answers to many important problems. However, when the problems require the evaluation of rare events, the number of simulation runs (or replications) required to achieve a reasonable confidence interval can be prohibitively high. There are two main approaches that have been used in the literature to improve the efficiency of rare-event simulations: importance sampling and splitting. The basic idea of splitting is to create separate copies of the simulation whenever the simulation gets close to the rare-event. Effectively, this multiplies promising runs that are “near” the rare-event, thus improving the likelihood of observing the rare-event.

In this project, we will develop a new rare-event simulation methodology for decision making. The key idea is to integrate the notions of optimal budget allocation and splitting to optimally allocate the limited computing budget so that the overall efficiency is maximized. For simulation analysis of a single system, we want to determine the optimal numbers of simulation runs among a number of splitting levels so that the variance of the rare-event estimator is minimized. For a comparison of multiple designs, we want to determine the optimal numbers of simulation runs for each splitting level in all designs so that the overall simulation efficiency is maximized. We believe that it is possible to achieve a speed-up that is more than the product of the individual speed-ups.  The proposed rare-event simulation method will enable decision makers to efficiently analyze the impact of disruption and quickly identify the best mitigation decision.