Stochastic Refinery Planning with Risk Management

Abstract

This work proposes a two-stage stochastic programming model with fixed recourse via scenario analysis with incorporation of risk management for an optimal midterm refinery planning that addresses three factors of uncertainties: prices of crude oil and saleable products (in the objective function), product demands (in the RHS coefficients), and product yields (in the LHS coefficients). Compensating slack variables and “discrepancy costs” are employed to explicitly account for constraints' violations to increase model tractability. Variance is adopted as the risk measure, with its shortcomings highlighted and mean-absolute deviation proposed as an improved alternative. A representative numerical example is illustrated.