Data-driven Process Systems Engineering Lab
Data-driven Process Systems Engineering Lab
2015 AIChE Annual Conference in Salon F (Salt Lake Marriott Downtown at City Creek)
November 10, 2015 at 12.30
Planning of the production of petrochemical products involves solving large-scale optimization models incorporating information such as: flowsheet topology, process capacity, operating constraints, intermediates and final product quality constraints, and market demand and costs [1-2]. Optimization of planning operations becomes significantly challenging when multiple refinery and chemical processing plants are integrated, creating complex networks and interactions through exchange of streams across plants . Despite the existence of rigorous mathematical models for specific unit operations involved in petrochemical operations, the models which are suitable for entreprise-wide planning optimization must maintain a balance between accuracy and complexity in order to create overall tractable planning optimization formulations. In the past, linear formulations have predominantly been utilized, however, it has been verified that linear models lead to suboptimal solutions reducing the profitability and product quality [1-2]. Undoubtedly, planning operations have non-linear behavior in terms of the yields and properties of the intermediate and product streams, due to nonlinearities in the operation of individual unit operations as well as blending of streams. Commercially available software developed for planning operations either use linear approximation models, while in cases where non-linear models are used identification of the global optimum is not guaranteed. The need for accurate non-linear models and their integration within industrial-scale custom-made superstructures to form Mixed-Integer Non-Linear Optimization (MINLP) problems coupled with global optimal optimization methods is one of the posed future challenges in the field [1-2].
Hybrid or data-based models have been used within smaller planning problems in the literature, due to their simplicity, low computational expense and ability to accurately capture the behavior of real industrial data which is available in abundance [4-5]. One of the disadvantages of data-based modeling is their limited extrapolating abilities, requiring updating in case the operation or raw material properties have changed significantly. In this work we present the development of an industrial-scale flowsheet superstructure comprised of a hybrid set of non-linear unit operation models (first-principles and data-based). We use this grey-box framework to accurately represent an integrated petrochemical plant comprised of one refinery and two ethylene cracking plants. We show the performance of our fully automated framework which performs: (a) automated updating of unit operation model parameters based on real plant data, (b) integration of unit operation models with blending and pooling operations and mass-balance and capacity constraints, and (c) global optimization of the formulated planning problem using ANTIGONE . Using the developed model, we solve multiple scenarios by modifying several uncertain parameters such as product demands, pricing, quality specifications and crude oil availability and properties and perform analysis on the obtained solutions. We also compare the obtained solutions with existing planning practices for the specific multisite plant and show that we have obtained significantly increased the profit while satisfying quality specifications.
1. N.K. Shah, Z. Li, M.G. Ierapetritou, Petroleum Refining Operations: Key Issues, Advances and Opportunities, Industrial Engineering & Chemistry Research, 2011, 50, 1161-1170.
2. I. E. Grossmann, Challenges in the Application of Mathematical Programming in the Enterprise-wide Optimization of Process Industries, Theoretical Foundations of Chemical Engineering, 2014, 48(5), 555-573.
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. V. Mahalec, Y. Sanchez, Inferential monitoring and optimization of crude separation units via hybrid models, Computers and Chemical Engineering, 2012, 45, 15-26.
5. W. Li, C. Hui, A. Li, Integrating CDU, FCC and product blending models into refinery planning, Computers and Chemical Engineering, 2005, 29, 2010-2028.
6. 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.