(201a) Big Data Analysis and Global Optimization for Petrochemical Planning Operations | AIChE

(201a) Big Data Analysis and Global Optimization for Petrochemical Planning Operations


Boukouvala, F. - Presenter, Texas A&M University
Li, J., Princeton University
Recently, the amount of collected data has experienced exponential growth, and the potential of using data for development of novel practical solutions is being realized by both academia and industry [1]. In this work, we have used high volumes of information-packed data in order to develop a data-driven planning model for an integrated refinery-petrochemical complex.  Planning problems can be formulated as mixed-integer non-linear optimization models coupling process models, mixing and pooling models, mass-balance and unit connection constraints, demand and quality constraints, as well as unit and production mode selection decisions [2-3]. Models used within large-scale planning problems must maintain a balance between accuracy and complexity in order to create overall tractable optimization problems. Simplified linear formulations have been used in the past; however, they lead to suboptimal solutions due to their inability to capture the inherent nonlinear behavior of process operations and interactions.

Through this work, we aim to demonstrate the power of using past assimilated data for the prediction of future optimal operating plans, when coupling data-driven concepts, with first-principle information and deterministic global optimization. The main objectives of this work include (a) the analysis and preprocessing of noisy industrial data from 43 units present in three connected plants (one refinery and two chemical processing plants); (b) the identification of the critical correlations between input flowrate streams and input stream properties with outlet yields and outlet stream properties; (c) the identification of the form of the input-output functional forms for each of the above correlations and the global optimization of their parameters based on a least-squares minimization objective; (d) the integration of the unit operation models through connection of input-output streams; (e) the integration of pooling and blending units, demand constraints, capacity constraints and operating cost constraints; and (f) the global optimization of the complete planning model for profit maximization.

The input-output models included in this formulation range from linear, to quadratic, to polynomial, and contain parameters which are globally optimized based on the data which spans two years of operation. For automated updating of the model parameters, we have developed a user-friendly computational platform which is used to update cost, price, demand and specification parameters for the planning period of interest. The global solution of the complete planning model is found using the global optimization solver ANTIGONE [4]. Using the developed framework, we have obtained results for multiple case studies and we have obtained solutions with up to 50% improvement in profit.    


1. S.J. Qin, Process Data Analytics in the Era of Big Data, AIChE Journal, 60(9), 3092-3100.

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

3. K. Al-Qahtani and A. Elkamel, Multisite Refinery and Petrochemical Network Design: Optimal Integration and Coordination, Industrial Engineering & Chemistry Research, 2009, 48, 814-826.

4. R. Misener and C.A. Floudas, ANTIGONE: Algorithms for Continuous/ Integer Global Optimization of Nonlinear Equations, Journal of Global Optimization, 2014, 59(2-3), 503-526.

Acknowledgments: The authors would like to acknowledge the Chinese Academy of Sciences Visiting Professorship for Senior International Scientists.