(26b) Deterministic Global Optimization Approach to Midterm Planning of an Industrial Integrated Petroleum-Petrochemical Facility

Authors: 
Guillén-Gosálbez, G., Imperial College London
Uribe-Rodriguez, A., Colombian Petroleum Institute (ICP), ECOPETROL S.A.
Chachuat, B., Imperial College London
Castro, P. M., Faculdade de Ciências, Universidade de Lisboa
Refining processes transform petroleum into a wide range of valuable products, from fuels to petrochemicals. Profitability of this industry relies on key activities such as crude oil evaluation, petroleum allocation, fuel blending and selection of optimal operational windows. Linear representations for refinery planning operations tend to under- or overestimate production rates and products quality, making questionable the effectiveness of optimization based approaches1–3. Recently, refining process accuracy has been improved via metamodeling4–6 and empirical correlations7. Also, complex petroleum supply chains can be represented by a 3-tuple composed of a mixer – process unit – splitter8 linked between them. Despite the notorious advantage of achieving a better process representation, the presence of non-convexities leads to obtain local or suboptimal solutions. In terms of strategies for addressing this problem, heuristic-based algorithms9,10 have been applied to find near optimal solutions, while deterministic global optimization methods3 ensure a mathematical global optimum. However, applying these approaches for a real-size problem is challenging owing to process integration between petroleum – petrochemical technologies, competition for intermediate raw materials and the multiple grades of commodities to produce. In addition, if process models are represented by non-linear expressions, the chance of finding a local solution is significant.

Herein, we formulate a Mixed-Integer Quadratically-Constrained Quadratic Program (MIQCQP) to solve within a certain global optimality gap a midterm supply-chain problem for a full-scale integrated refinery – petrochemical complex in the context of the Colombian hydrocarbon market. Our approach clusters the refinery topology into subprocesses according to their functionality. For each cluster, a relaxed model (MILP) based on generalized11 and piecewise McCormick envelopes is formulated. The number of partitions (NP) for the discretized variables inside the cluster change dynamically, while variables that do not belong to the cluster are not discretized. The MILP solution provides a Lower Bound (LB) for the original problem. Then, fixing the binary variables obtained from the relaxed model solution, the original MIQCQP is transformed into a quadratic problem (QP) which provides an Upper Bound (UB). If the UB is improved, Optimality-Based Bound Tightening (OBBT)12 is applied to reduce the domain for the variables belonging to the cluster. This procedure is repeated until a stopping criteria is met, such us reaching maximum runtime, exploring all clusters or obtaining and optimality gap less than epsilon.

The methodology was tested through five cases studies that recreate typical planning scenarios for the Colombian hydrocarbon industry. The model has about 6975 equations, 35104 nonlinear terms derived from bilinear and trilinear expressions, 9592 and 279 continuous and discrete variables respectively. Results show that commercial solvers for deterministic global optimization13,14 get stuck at a local optimum solution with an optimality gap above 50%, on average. Furthermore, there is no improvement in the UB if the CPU time is increased from 2 to 6 hours. In contrast, we found a better UB for all the case studies with a maximum runtime of 1.3 hours on average. This demonstrates that clustering decomposition is a promising solution strategy for problems of that scale. Future work will incorporate the Reformulated Normalized Multiparametric Disaggregation Technique15 for comparison with the piecewise McCormick envelopes.

References

  1. Khor CS, Varvarezos D. Petroleum refinery optimization. Optim Eng. 2017;18(4):943-989. doi:10.1007/s11081-016-9338-x
  2. Alattas AM, Grossmann IE, Palou-Rivera I. Refinery production planning: Multiperiod MINLP with nonlinear CDU model. Ind Eng Chem Res. 2012;51(39):12852-12861. doi:10.1021/ie3002638
  3. Castillo Castillo P, Castro PM, Mahalec V. Global optimization algorithm for large-scale refinery planning models with bilinear terms. Ind Eng Chem Res. 2017;56(2):530-548. doi:10.1021/acs.iecr.6b01350
  4. Li W, Hui CW, Li A. Integrating CDU, FCC and product blending models into refinery planning. Comput Chem Eng. 2005;29(9):2010-2028. doi:10.1016/j.compchemeng.2005.05.010
  5. Alhajri I, Elkamel a., Albahri T, Douglas PL. A nonlinear programming model for refinery planning and optimisation with rigorous process models and product quality specifications. Int J Oil, Gas Coal Technol. 2008;1(3):283. doi:10.1504/IJOGCT.2008.019846
  6. Guerra OJ, Le Roux GAC. Improvements in petroleum refinery planning: 1. Formulation of process models. Ind Eng Chem Res. 2011;50(23):13403-13418. doi:10.1021/ie200303m
  7. Alattas AM, Grossmann IE, Palou-Rivera I. Integration of nonlinear crude distillation unit models in refinery planning optimization. Ind Eng Chem Res. 2011;50(11):6860-6870. doi:10.1021/ie200151e
  8. Neiro SMS, Pinto JM. A general modeling framework for the operational planning of petroleum supply chains. Comput Chem Eng. 2004;28(6-7):871-896. doi:10.1016/j.compchemeng.2003.09.018
  9. Andrade T, Ribas G, Oliveira F. A Strategy Based on Convex Relaxation for Solving the Oil Refinery Operations Planning Problem. Ind Eng Chem Res. 2016:acs.iecr.5b01132. doi:10.1021/acs.iecr.5b01132
  10. Guerra OJ, Uribe-Rodriguez A, Montagut SM, Duarte LA, Angarita JD. A Solution Strategy for Large-Scale Nonlinear Petroleum Refinery Planning Models. In: AIChE Annual Meeting. Salt Lake city; 2010. https://www.aiche.org/conferences/aiche-annual-meeting/2010/proceeding/p.... Accessed August 3, 2018.
  11. McCormick GP. Computability of global solutions to factorable nonconvex programs: Part I - Convex underestimating problems. Math Program. 1976;10(1):147-175. doi:10.1007/BF01580665
  12. Puranik Y, Sahinidis N V. Domain reduction techniques for global NLP and MINLP optimization. Constraints. 2017;22(3):338-376. doi:10.1007/s10601-016-9267-5
  13. Sahinidis N V. BARON: A general purpose global optimization software package. J Glob Optim. 2004;8(2):201-205. doi:10.1007/bf00138693
  14. Misener R, Floudas CA. ANTIGONE: Algorithms for coNTinuous / Integer Global Optimization of Nonlinear Equations. J Glob Optim. 2014;59(2-3):503-526. doi:10.1007/s10898-014-0166-2
  15. Andrade T, Oliveira F, Hamacher S, Eberhard A. Enhancing the normalized multiparametric disaggregation technique for mixed-integer quadratic programming. J Glob Optim. 2018;73(4):701-722. doi:10.1007/s10898-018-0728-9
Topics: