Efficient Two-Level Hybrid Algorithm for the Refinery Production Scheduling Problem Involving Operational Transitions

详细信息    查看全文

• 作者：Lu Zhang ; Yongheng Jiang ; Xiaoyong Gao ; Dexian Huang ; Ling Wang
• 刊名：Industrial & Engineering Chemistry Research
• 出版年：2016
• 出版时间：July 20, 2016
• 年：2016
• 卷：55
• 期：28
• 页码：7768-7781
• 全文大小：746K
• 年卷期：0
• ISSN：1520-5045

文摘

An overall refinery production scheduling problem involving operational transitions was studied by Shi et al. (Ind. Eng. Chem. Res. 2014, 53 (19), 8155–8170), which is an intractable large-scale mixed-integer linear programming (MILP) problem. To deal with this challenge, an efficient two-level hybrid algorithm based on the hierarchy of decisions is proposed. In the outer level, the key techniques of a discrete particle swarm optimization algorithm are designed to search for the operating states assignment of production units. A queue-based solution representation is offered by an elaborate encoding scheme for the outer-level problem so that much fewer discrete decision variables are involved. All the constraints on operating states assignment are represented with the encoding scheme and can be satisfied by all the particles throughout the evolutionary process. In the nested inner level, under the assignment specified in the outer level, the detailed production process is optimized by dual simplex method, which includes the production, consumption, inventories of materials, and deliveries of products. Considering that Zhang et al. (Ind. Eng. Chem. Res. 2015, 54 (32), 7871–7889) studied a similar problem and provided an improved MILP model with much better performance, computational comparisons are made between the two-level hybrid algorithm and a full-space MILP model which is mainly adapted from the work of Zhang et al. The results show that the proposed method takes much less time to obtain near-optimal or even optimal solutions robustly. The advantage of the hybrid algorithm is especially obvious to large-scale problems, and better solutions can be obtained in minutes when GAMS cannot reach the optimality within reasonable times.