Supply Chain Optimization for Refinery with Considerations of Operation Mode Changeover and Yield Fluctuations

Abstract

Stochastic programming is employed to achieve optimization for the multiperiod supply chain problem in a refinery with multiple operation modes under the uncertainty of product yields. With dramatic fluctuations of product yields at the beginning of operation mode changeover, the product yields tends to stabilize after the changeover is finished. Markov chain is utilized here to describe the dynamic characteristic of product yield fluctuations. The distribution of yield fluctuation in each period is usually unknown since it depends on the decision variable of operation mode changeover. Therefore, the resulting chance constrained programming is more complicated than general situations where the distribution characteristic of stochastic variable is known in each period. This problem can be solved by the big-M method and by transforming chance constrained inequalities into a group of equivalent deterministic inequalities. This method provides a universal approach for similar chance constrained programming in which the distribution of stochastic variable depends on binary decision variables. Case studies show that the proposed modeling and solving approach can provide an effective decision-making guidance that balances confidence level and economic interests for supply chain optimization problems with multiple operation modes under yield uncertainty.

Planning under Demand and Yield Uncertainties in an Oil Supply Chain

“…This character is closely related to the changeover of operation modes. This work is an extension of Yang, Gu, and Rong’s work …”

“…It should be noted that, in iteration 5, there is only one simulation needed, because the operation mode list is the same as the one in iteration 4. After six iterations, the best simulation result is in the sixth iteration where E(Cost) = $8.5193 Â 10 9 and the corresponding objective value is $8.5003 Â 10 9 .…”

“…For instance, product yield will be dramatically influenced by operation mode changeover. Yang 9 and co-workers assumed that the product yield is only influenced by its last state, which has the Markov property being time-independent and time-homogeneous. A constraint programming approach was employed to keep the inventory over safety stock level within the largest probability.…”

“…Jonsbraten et al 12 proposed an implicit enumeration algorithm addressing such uncertainty. In Yang's model, 9 product yield probability distribution was generated when all the available mode changeover list was enumerated. Goel and Grossmann 10,13 Industrial & Engineering Chemistry Research ARTICLE dealt with the gas field development problem under uncertainty in size and quality of reserves where decisions on the timing of field drilling yield an immediate resolution of the uncertainty.…”

CVaR-Based Approximations of Wasserstein Distributionally Robust Chance Constraints with Application to Process Scheduling

“…Even with a small number of uncertain variables, solving the above optimization problem is computationally intensive. Although Li et al, − Rong et al, , Yang et al, , and Ostrovsky et al , have derived multivariate integration approaches and solution algorithms for CCP, none of them is universally applicable to all commonly used probability distributions. There are merely a few special cases under which chance constraints can be reformulated into tractable convex constraints.…”