Petroleum refining and petrochemical industries are facing more challenges than ever before. Crude oil price is continuously rising while the quality is becoming low. Although the products' price can be increased to some extent, however, the major petroleum products, such as gasoline and diesel, have to meet more rigorous quality specifications at the same time. To efficiently make profits under these challenges, faster and more accurate planning and scheduling models for both incorporation and plant levels are urgently needed. In this paper, an integrated planning model for maximum profit making has been developed. The internal interactions between incorporation and plant wide models are also discussed.

A plant-level planning model forecasts detailed raw material requirement, process and product management in the next few months. It is a multi-period mixed-integer nonlinear model. Multi-period means future plan is divided into several time intervals. The intervals in the near future describe detailed plans in a short time period to cooperate with scheduling. The intervals in the far future describe brief plans in a long time period. Mixed-integer features discrete information including unit minimal capacity requirements, crude oil shipping constraints and utility constraints. Nonlinearity suggests quality related information, which is including swing cut, whole plant quality tracking and nonlinear blending. The corporation-level planning model is the integration of over 20 plant-level models plus distribution network. It predicts corporation's crude oil trading, distribution of raw materials, intermediate products and final products.

Both corporation-level and plant-level models should be solved very quickly. Usually, users want to get 6 solution scenarios within one day, which suggests each scenario should be completed within 1 hour. Given averagely 5 times tuning for one scenario, a corporation-level model should be solved in 5~7 minutes, which is too stringent. Thus, plant-level planning models should be simplified before the integration. To avoid the possible discrepancy between corporate and plant level models, a simplified methodology imbedded with many methods and heuristic rules is presented in this paper.

The proposed bi-level planning model system has been applied in a large petroleum corporation in Asia and has brought tremendous profit.