Management of Data Uncertainty in Dynamic Programming

Summary

This chapter introduces an innovative approach to Dynamic Programming optimization to deal with data uncertainty problem. The proposed approach incorporates confidence measure to quantify the data uncertainty. The concept of confidence level enables the utilization of alternative data sources for estimating uncertain data. This algorithm of fuzzy optimization has been applied to optimize the availability of unique spare parts of a power station. In particular, the paper describes two types of confidence level propagation: confidence level at the component level and confidence level at the power station level. Simulated case studies have been included to show the characteristics of the fuzzy aggregators and their suitability in propagating the confidence level. Empirical research on fuzzy aggregators is useful because it allows different approaches (from conservative to risky) to optimum decision making. This notion of dynamic decision making is addressed in this chapter. Although the study implements Dynamic Programming in optimizing the spare parts inventory of an electrical power plant, the proposed technique is generally applicable to optimize similar categories of objective functions when fuzzy aggregators can be determined by empirical research.