Gasoline blending is an important downstream operation in the refinery. The operation is susceptible to uncertainties such as fluctuation in component quality, fluctuation in demand, and a combination of both. These problems naturally involve multiple objectives and non-linear terms corresponding to the mixing of the components for which the genetic algorithm-based approach is more suitable compared to traditional mathematical programming. In this study, such graphical genetic algorithm-based reactive scheduling approach is developed which can handle dynamic changes in component quality and demand as an additional layer of decision making over the nominal scheduling. Three industrial-scale examples are solved using the developed approach for both single- and two-objective optimizations while handling the uncertainty of 10% increase in demand and 5% decrease in component quality. In single-objective optimization, the production cost is minimized whereas in two-objective optimization additionally the fluctuation in blending processing rate is minimized.

汽油混合是炼厂的重要下游操作。操作容易受到不确定性的影响，例如组件质量的波动，需求的波动以及二者的结合。这些问题自然涉及多个目标和对应于组分混合的非线性项，与传统的数学编程相比，基于遗传算法的方法更适合这些组分的混合。在这项研究中，开发了基于图形遗传算法的反应式调度方法，该方法可以处理组件质量和需求的动态变化，作为名义调度之外的决策层。使用已开发的方法对单目标和两个目标进行了优化，解决了三个工业规模的示例，同时处理了需求增加10％和组件质量减少5％的不确定性。在单目标优化中，生产成本降至最低，而在二目标优化中，混合加工速率的波动也降至最低。