Abstract
Optimization of process technologies under uncertainty has been extensively studied in the literature. It provides a rigorous and powerful tool for the design of advanced technologies. Two methods are available for optimization of process models under uncertainty, which are stochastic optimization and stochastic programming. From the results of the two methods, Expected Value of Perfect Information (EVPI) can be estimated, which provides decision-makers the expected value of maximum benefit of reducing uncertainty. However, optimization of process models under uncertainty has not made distinctions between variability and uncertainty. Variability is a heterogeneity of values for a quantity over time, space or among different members of a population, while uncertainty is a lack of information. This study proposes two methodologies for optimization of process models when both variability and uncertainty in model parameters are considered. One is a coupled stochastic optimization and programming method, which involves stochastic optimization for each realization of variability and enables one to evaluate the effect of uncertainty on optimal designs. The other one is a two-dimensional stochastic programming technique, which features stochastic programming for each realization of variability and produces two-dimensional distributions of deterministic optimal solutions. Comparing the outputs of the two methods, both point estimates and confidence intervals of EVPI can be estimated. The two methods are demonstrated through application to optimization of the cost and emissions of a KRW-Gasifer based IGCC system when both variability and uncertainty in model parameters are considered.
The methodologies proposed and demonstrated in this study are helpful to design and evaluation of advanced technology applications where cost minimization, risk analysis, environmental compliance and R&D priority remain important issue.
|
