(604g) Effective Global Optimization Framework Coupled with Robust Optimization for Refinery-Wide Planning Operations Under Uncertainty | AIChE

# (604g) Effective Global Optimization Framework Coupled with Robust Optimization for Refinery-Wide Planning Operations Under Uncertainty

Conference

Year

Proceeding

Group

Session

Time

## Authors

Tsinghua University
Tsinghua University
Optimization for the refinery planning is of vital importance for the desirable profitability of a petroleum refinery[1]. Traditional commercial approaches for refinery planning (i.e. PIMS[2], RPMS[3]) often correlate input and output variables of processing units with linear models for yield and property prediction purpose. However, non-rigorous linear models may not only reduce the overall profitability or product quality but also fail to provide the unitsâ€™ operation conditions. Quite a few literatures have focused on the planning operations with empirical models for crude distillation columns, Fluid Catalytic Cracking units et al[4-6]. On the other hand, fluctuations on the demand and price of products also affect significantly the overall profitability. Therefore, we propose an effective modeling and global optimization framework coupled with robust optimization for refinery-wide planning operations under demand/price uncertainties

For modeling the planning operations for the entire refinery facility, swing-cut theory is implemented for modeling crude distillation units; while the explicit nonlinear surrogate models, which are developed by ALAMO[7], are employed for predicting the yield and property of products for all other secondary processing units based on the industrial data sets. Operation conditions such as temperature, pressure, and specific gravity are included as the inputs for surrogate models with the forms of bilinear, quadratic terms, or root squared terms. Based on the extensive observations on the demand and price variations in China, â€œinterval+ellipsoidalâ€ uncertainty set is employed for the uncertainty distributions[8]. Therefore, a noncovex mixed integer nonlinear programming (MINLP) is formulated.

In order to solve the formulated MINLP model to global optimality with reasonable time, a tailored global optimization algorithm is developed to seek the highest profitability. Taking the advantages that the number of auxiliary binary variables and constraints of relaxed MINLP model size can be reduced by the enhanced normalized multiparametric disaggregation (ENMDT) [9]when compared to the classical NMDT relaxation[10], a convex MINLP subproblem is generated from the original MINLP through the relaxation of bilinear terms via ENMDT. Fixing the binary variables obtained from the convex MINLP solution, the original MINLP is subsequently transformed to a nonlinear programming problem (NLP) which provides a lower bound. The convex MINLP-NLP iteration is repeated until the solution meets the stop criterion. Optimality-based bound tightening (OBBT) is also implemented to tighten the lower and upper bounds during the iterations.

A large-scale industrial case study is solved to examine the efficiency of our developed nonlinear modeling and the associated global optimization approach. The relationship between the profitability and the shape of the uncertainty set is revealed. Computational results indicate that the developed methodology achieves a higher profitability while satisfying quality specifications when compared to the classical approaches which fix the product yields for secondary processing units. It also demonstrates that the proposed global optimization algorithm outperforms the state-of-the-art commercial solvers such as BARON[11]and ANTIGONE[12].

keywords: Planning operations; global optimization; Nonconvex MINLP; robust optimization; Surrogate Model

Reference:

1. Shah NK, et al. Petroleum refining operations: Key issues, advances and opportunities, Industrial Engineering & Chemistry Research, 2011; 50: 1161-1170.
2. RPMS (Refinery and Petrochemical Modeling Systems): A System Description; Bonner and Moore: Houston, TX, 1979.
3. Aspen PIMS System Reference, Aspen Technology Inc.; Burlington, MA 2010.
4. Li WK, et al, Integrating CDU, FCC, and product blending models into refinery planning, Computers Chemical Engineering, 2005; 29: 2010-2028.
5. Alattas AM, et al. Integration of nonlinear crude distillation unit models in refinery planning optimization. 2011; 50: 6860-6870.
6. Li J, et al. Data-driven mathematical modelling and global optimization framework for entire petrochemical planning operations. AIChE Journal. 2016; 62: 3020-3040.
7. Cozad A. Learning surrogate models for simulation-based optimization. AIChE Journal. 2014; 60: 2211-2227.
8. Li ZK, Floudas CA. A Comparative Theoretical and Computational Study on Robust Counterpart Optimization: I. Robust Linear Optimization and Robust Mixed Integer Linear Optimization. Industrial Engineering & Chemistry Research, 2011; 50: 10567-10603.
9. Andrade T, et al. Enhancing the normalized multiparametric disaggregation technique for mixed-integer quadratic programming. Journal of Global Optimization. 2019; 73: 701-722.
10. Castro PM. Normalized multiparametric disaggregation: an efficient relaxation for mixed-integer bilinear problems.Journal of Global Optimization. 2016; 64: 765-784.
11. Tawarmalani M, Sahinidis NV. A polyhedral branch-and-cut approach to global optimization. Mathematical Programming. 2005; 103: 225-249.
12. Misener R, Floudas CA. ANTIGONE: Algorithms for coNTinuous / Integer Global Optimization of Nonlinear Equations. Journal of Global Optimization. 2014; 59: 503-526.