(703c) Medium-Term Scheduling of Integrated Gasoline Blending and Delivery Operations Using Enhanced Rolling Horizon Decomposition Approach

Authors: 
Li, J., The University of Manchester
Floudas, C. A., Texas A&M University
Xiao, X., Institute of Process Engineering, Chinese Academy of Sciences
Gasoline is one of the most profitable products of an oil refinery and can account for as much as 60-70% of total revenue [1-3]. Several gasoline cuts or fractions from various processes are typically blended together to produce different gasoline grades in order to meet customer orders of varying specifications and delivery dates. The entire operations involve handling a large numbers of orders, delivery dates, blenders, blend components, product tanks, and quality specifications, making scheduling highly complex. Optimal scheduling of such blending and delivery operations has several advantages such as avoiding ship demurrage, improving customer satisfaction, minimizing quality give-aways, reducing transitions, and slop generation, exploiting low-quality cuts, and reducing inventory costs. However, mathematical modeling of blending operations often introduces non-convex nonlinear terms, making the problem a large-scale complex non-convex mixed-integer nonlinear programming (MINLP). Scheduling based on industrial experiences or heuristics often leads to inferior schedules with costly quality give-aways, many transitions, and order demurrage. Thus, it is imperative to seek advanced techniques such as mixed-integer programming approach for simultaneous treatment of recipe, blending, scheduling and delivery operations.

The early work [4-5] mainly focused on finding optimal blending recipes of various intermediate fractions and some additives to meet product quality specifications. Scheduling decisions such as resources allocations and temporal decisions were not considered [2]. Although several efforts[6-8] have developed models and solution approaches to address the problem of scheduling of gasoline blending operation, an integrated treatment of recipe, blending, scheduling, and delivery is missing [2]. Li et al. [2], Li and Karimi [9], Castillo-Castillo and Mahalec [10-12], Li et al. [13] and Cerda et al. [14] addressed the problem of integrated gasoline blending and delivery operations. They incorporated many problem features such as non-identical parallel blenders, minimum run length and amount in a blend run, constant blending rates, changeover, one blender charging at most one tank at a time, multi-purpose product tanks, piecewise constant profiles for blend component qualities and feed rates, etc. It should be mentioned that recently Li et al. [13] did improve MILP relaxation of Li and Karimi [9] and proposed a hybrid global optimization approach to solve the MINLP model to e- (e.g., 1%) global optimality. All of these models and solution algorithms are restricted to short-term scheduling. When the scheduling horizon increases to medium term, they may need much more computational efforts or fail to find a feasible solution.

To address the problem of medium-term scheduling of gasoline blending and delivery operations, Castillo-Castillo and Mahalec [15] developed a multi-period inventory pinch-based algorithm to solve continuous-time scheduling models (MPIP-C algorithm). The entire problem was decomposed into three-level subproblems. The top-level model computed optimal recipes for aggregated blends over periods initially delineated by inventory pinch points. The second level model used fixed blend recipes to compute an approximate schedule. The last level model was the short-term continuous-time model that generated detailed schedules for the entire problem.

In this presentation, we use the model of Li et al. [13] and enhance rolling horizon decomposition approach to address the problem of medium-term scheduling of gasoline blending and delivery operations. The entire problem is decomposed into two-level sub-problems (i.e., Level-1 and Level-2 models). While the Level-1 model is utilized to determine the current time horizon and corresponding orders that should be included into the current horizon, the level-2 model is the same short-term scheduling model of Li et al. [13] to determine the detailed scheduling operations in the current horizon and minimize the total operating cost. We first solve 14 industrial-scale examples from Li et al. [13] and compare with those of Li et al. [13]. Then, we solve the examples from Castillo-Castillo and Mahalec [15] and compare with their solution algorithm MPIP-C [15]. The computational results demonstrate that we can generate better solutions using our enhanced rolling horizon decomposition than those of MPIP-C algorithm.

References

[1] Jia, Z. Y.; Ierapetritou, M. Mixed-Integer Linear Programming Model for Gasoline Blending and Distribution Scheduling. Industrial and Engineering Chemistry Research 2003, 42, 825-835.

[2] Li, J.; Karimi, I. A.; Srinivasan, R. Recipe Determination and Scheduling of Gasoline Blending Operations. AIChE Journal 2010, 56, 441-465.

[3] Shah, N. K.; Li, Z.; Ierapetritou, M. G. Petroleum Refining Operations: Key Issues, Advances, and Opportunities. Industrial and Engineering Chemistry Research 2011, 50, 1161-1170.

[4] Dewitt, C. W; Lasdon, L. S.; Waren, A. D.; Brenner, D. A.; Melhem, S. A. OMEGA: An Improved Gasoline Blending System for Texaco. Interfaces 1989, 19, 85-101.

[5] Rigby, B.; Lasdon, L. S.; Waren, A. D. The Evolution of Texacoâ??s Blending Systems: From OMEGA to Starblend. Interfaces 1995, 25, 64-83.

[6] Mendez, C. A.; Grossmann, I. E.; Harjunkoski, I.; Kabore, P. A. Simultaneous Optimization Approach for Off-Line Blending and Scheduling of Oil-Refinery Operations. Computers and Chemical Engineering 2006, 30, 614-634.

[7] Glismann, K.; Gruhn, G.; Short-Term Scheduling and Recipe Optimization of Blending Processes. Computers and Chemical Engineering 2001, 25, 627-634.

[8] Joly, M; Pinto, J. M. Mixed-Integer Programming Techniques for the Scheduling of Fuel Oil and Asphalt Production. Trans IChemE. Part A. 2003, 81, 427-447.

[9] Li, J.; Karimi, I. A. Scheduling Gasoline Blending Operations from Recipe Determination to Shipping using Unit Slots. Ind. Eng. Chem. Res. 2011, 50, 9156-9174.

[10] Castillo, P. A. C.; Mahalec, V. Inventory Pinch Based, Multiscale Models for Integrated Planning and Scheduling-Part I: Gasoline Blend Planning. AIChE Journal 2014, 60, 2158-2178.

[11] Castillo, P. A. C.; Mahalec, V. Inventory Pinch Based, Multiscale Models for Integrated Planning and Scheduling-Part II: Gasoline Blend Scheduling. AIChE Journal 2014, 60, 2475-2497.

[12] Castillo-Castillo, P. A.; Mahalec, V. Improved Continuous-Time Model for Gasoline Blend Scheduling. Computers and Chemical Engineering 2016, 84, 627-646.

[13] Li, J.; Xiao, X.; Floudas, C. A. Integrated Gasoline Blending and Order Delivery Operations: Part I. Short-Term Scheduling and Global Optimization for Single and Multi-Period Operations. AIChE Journal 2016, 62, 2043-2070.

[14] Cerda, J.; Pautasso, P. C.; Cafaro, D. C. A Cost-Effective Model for the Gasoline Blend Optimization Problem. AIChE Journal 2016, In press. DOI: 10.1002/aic.15208.

[15] Castillo-Castillo, P. A.; Mahalec, V. Inventory Pinch Gasoline Blend Scheduling Algorithm Combining Discrete- and Continuous-Time Models. Computers and Chemical Engineering 2016, 84, 611-626.