| 초록 |
Operational planning for crude oil procurement and refinery optimization is studied considering a number of different crude oil types, a network of processing units, and multiple products of desired specs. The main complication is that uncertainties exist on the prices of crudes and products. In addition, to ensure optimal inputs into the refinery at all times under constantly changing the prices, one needs to manage the inventories of crude oils. In this study, the marginal value of storing each crude at given price state is approximated by solving the Markov decision process (MDP) that is formulated with considering recurring dynamics of crude inventory and decisions under uncertainties on price. In the MDP formulation, one-day reward and inventory transition are computed by refinery optimization model, which is formulated as two-stage stochastic programming where the first stage is crude imports followed by product exports with price uncertainty realization. The performance of the refinery operation with the evaluated value of crude inventory is verified through a numerical case study. |
