Risk Informed Nuclear Asset Management - Models and Methods

University Of Texas At Austin, Austin TX

Investigators

Abstract

This grant provides funding to solve an important class of stochastic optimization models in which the governing probability distributions change according to decisions made over time. Specifically, this problem will be addressed for single- and multi-item maintenance optimization models as well as executive-level management decisions involving allocating funds across a collection of projects. First, a method of adaptively updating weights on a mixture of failure distributions in the context of a stochastic dynamic program will be developed. Then, an infinite-dimensional hierarchical Bayes model will be constructed by transforming the problem to a finite state space using a novel implementation of existing Markov Chain Monte Carlo methods called Polya trees. The resulting models will be solved by stochastic dynamic programming or stochastic integer programming, as appropriate. Sampling-based methods will be incorporated, as necessary. The maintenance optimization and executive decision-making problems will be formulated, and the solutions implemented, jointly with the Risk Management Group of the South Texas Project Nuclear Operating Company (STPNOC). If successful, solutions to both the higher-fidelity single-item maintenance models and the multi-item maintenance models can positively affect financial performance measures, as well as production and safety goals at STPNOC and other nuclear power companies. The proposed models and methods may be distributed more widely through the nuclear power industry due to collaboration with the Electric Power Research Institute.