(577l) Eliminating Gaps Between Advanced Planning and Scheduling Models

Advanced planning and scheduling systems are widely used in oil refining and petrochemical plants all over the world. A planning system is focused on corporation-wide or plant-wide optimization, and gets production plan that is relatively preliminary. A scheduling system is focused on plant-wide or unit-wide realization, and gets detailed schedule schemes based on the plan. A plan and the corresponding schedule should be consistent, otherwise a gap between them appears and either the plan or the schedule should be rebuilt. This paper discussed when will the gap between planning and scheduling be generated and how to eliminate the gap.

Primary reason for the gap is the difference of periods. A production plan is normally involved in one or more periods of year, month or week. A schedule is involved in periods of day or hour. A plan is “feasible” means it is feasible in monthly or weekly average, but it may not feasible in a specific day or hour. Many cases were reported that plans caused nonsensical scheduling requirements. For example, two streams may be pooled in a plan, but the two streams are produced in different time and no inventory available. To avoid such situation, common model structures were presented, which could be adopted by planning and scheduling models. For the planning model, special mixed-integer features were introduced to avoid periodic inconsistent between planning and scheduling. For the scheduling model, multi-period predictive features are introduced, which enable the scheduling model simultaneously to generate “near future” schedule and to predict “far future” plan. The “far future” plan produced by the scheduling system should be aligned with the one produced by the planning system, which will reduce the gap.

The methods have been used in several refineries. In a case study, plant-wide planning and scheduling models were built to deal with diesel transition problem through summer to winter. The planning model had 6 periods to optimize production plans from September to the next February, and the scheduling models were 35 periods rolling models. In this case, gap between planning and scheduling was effectively eliminated.