This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5)."
The proposed research is addressing the important question of the impact of human intervention on uncertainties that drive a particular system. Methods will be developed to assess whether interactions between decisions and uncertainties exist. If the answer is affirmative a new stochastic process equipped with "decision-dependent-prior distributions" is proposed. New models and Markov Chain Monte Carlo estimation procedures are defined to enable Bayesian inference for decision-dependent models. The second important question addressed is how to make optimal decisions in such environments. It has been approached by several researchers but with limited advances. This proposal provides a new procedure for tackling these very difficult optimization problems.
The proposed methods and algorithms will be applied to evaluate and construct maintenance system at nuclear power plants. Properly accounting for and quantifying uncertainty in maintenance policy decisions will reduce the variability in nuclear power production of electricity, while providing utility engineers quantified assessments of equipment performance on safety. Properly understanding and modeling the interplay between decisions and uncertainties, and their impact on the stochastic optimization problem at hand will enable utility engineers to better operate their power plants, while, simultaneously, minimizing costs. The potential impact is not limited to the electric power industry. In fact, the results will be applicable to any system in which future stochastic behavior is influenced by current and past human decisions. There are a plethora of real world applications that could merit from understanding and modeling decision-dependency along the lines of the research detailed in this proposal.
