Robust MINLP Optimization Model for Petrochemical Network Design using a Hybrid Stochastic-Probabilistic Approach
Systematic methods and tools for managing the complexity
Process Simulation & Optimization - III (T4-9c)
Keywords: Petrochemical network design, Strategic planning, Robust optimization
The Petrochemical industry is a network of highly integrated production processes. The products of one plant may have an end use but may also represent raw materials for another process as well. Most chemicals can be produced by many different sequences of reactions and production processes. This multiplicity of production schemes offers the sense of switching between production methods and raw materials utilization. This flexibility in petrochemical products production and the availability of many process technologies require adequate strategic planning and a comprehensive analysis of all possible production alternatives.
In this paper, we will develop a robust hybrid stochastic-probabilistic mixed-integer nonlinear programming (MINLP) model that accounts for uncertainties in product demand, product prices and process yields. The proposed approach presents an adequate and flexible handling of the different uncertainty types in the objective function, technological matrix coefficients, and left hand side parameters. Process yields uncertainty will be modeled as probabilistic (chance) constraints. This approach presents a realistic treatment of the uncertainty in process yields as it is difficult to assess the cost and benefits of second-stage decisions in case stochastic approach with recourse is adopted. Furthermore, chance constraints maintains the model size as apposed to the stochastic scenario based approach and therefore avoids the additional increase in model size. On the other hand, demand uncertainty will be modeled as a two-stage stochastic with recourse in order to maintain the flexibility of penalizing the shortfall and surplus costs and benefits of demand. Risk measures will be incorporated through a mean-absolute deviation approach in order to provide a more robust analysis of the petrochemical industry network. The deterministic MILP model will be based on the work of Al-Sharrah; Alatiqi; Elkamel; & Alper (2001).
See the full pdf manuscript of the abstract.
Presented Wednesday 19, 11:00 to 11:20, in session Process Simulation & Optimization - III (T4-9c).
