Optimal design of ethylene and propylene coproduction plants with generalized disjunctive programming and state equipment network models

Abstract

In this work, we propose a superstructure optimization approach for the optimal design of an ethylene and propylene coproduction plant. We formulate a superstructure that embeds ethane and propane steam cracking technologies, propane dehydrogenation and olefin metathesis processes. We represent the superstructure with a Generalized Disjunctive Programming model, and solve the problem through a custom implementation of the Logic-based Outer Approximation algorithm in GAMS. We propose a stateequipment-network representation to model potential distillation trains, as well as alternative acetylene reactor configurations. Rigorous models are formulated for distillation columns, compressors, turboexpanders, vessels and several process equipment units. The objective function is to maximize the net present value. We analyze four international price scenarios for raw material and utility costs, while considering global ethylene and propylene prices. We obtain the optimal scheme for each case. Numerical results show that the combination of ethane steam cracking, olefin metathesis and ethylene dimerization is the most profitable configuration under low ethane price scenarios, whereas the combination of ethane and propane steam cracking together with propane dehydrogenation is the optimal solution when the propane price is on the order of ethane price.

Equation-Oriented Optimization Applied to the Optimal Design of Carbon Capture Plants Using Rigorous Models

“…20,21 Currently, research and development teams use equationoriented packages to design, optimize, validate models, and address debottlenecking problems. 22,23 Exact derivatives are generally employed, and consequently, cheap sensitivity information is available at the optimal solution. In this approach, roundoff errors in gradients do not cause degradation in the performance of the solution strategy, resulting in a suitable tool for complex flowsheets with nested recycles and several design specifications.…”

“…In contrast, equation-oriented (EO) optimization relies on Newton-based solvers to achieve feasibility and to satisfy the optimality conditions (KKT) . Due to the advances of nonlinear programming algorithms, large-scale problems (>100,000 constraints and variables) can be solved efficiently using EO formulations while optimizing problems of this scale can be intractable using simulation-based optimization. , Currently, research and development teams use equation-oriented packages to design, optimize, validate models, and address debottlenecking problems. , Exact derivatives are generally employed, and consequently, cheap sensitivity information is available at the optimal solution. In this approach, roundoff errors in gradients do not cause degradation in the performance of the solution strategy, resulting in a suitable tool for complex flowsheets with nested recycles and several design specifications .…”

Homotopy continuation enhanced branch and bound algorithms for strongly nonconvex mixed‐integer nonlinear optimization

“…For example, the OA algorithm 15 with some modifications 40–42 has been applied to optimization of both single distillation column 41,43,44 and distillation sequence 45 modeled by rigorous material balance, phase equilibrium, summation and enthalpy balance (MESH) equations, synthesis of heat exchanger network, 4 simultaneous process and heat exchanger network optimization, 46 and simultaneous topology and parameter optimization of a multiple cantilever beam 42 and roller and sliding hydraulic steel gate 47 . The logic‐based OA algorithm 22 has also been applied to solve the synthesis of distillation sequence and optimize the whole plant using MESH model for distillation 48–50 . The nonlinear B&B has been applied to integrated design and control, 51,52 simultaneous scheduling and control, 53 process synthesis incorporating MESH models for distillation columns 54 .…”

“…47 The logic-based OA algorithm 22 has also been applied to solve the synthesis of distillation sequence and optimize the whole plant using MESH model for distillation. [48][49][50] The nonlinear B&B has been applied to integrated design and control, 51,52 simultaneous scheduling and control, 53 process synthesis incorporating MESH models for distillation columns. 54 However, the difficulty in solving nonconvex NLP subproblems may still lead to worse solutions or even no solution at all for the nonconvex MINLP problems, 52 especially when some accurate yet complex models are included.…”

Process schemes of ethanol coupling to C4 olefins based on a genetic algorithm for back propagation neural network optimization

“…Supported by experimental data, mathematical models can be used to predict the variation of the yield with other conditions in a multi-dimensional and dynamic manner, revealing the relationship between selectivity and catalyst types and operating parameters while providing some guidance for further experimental research and industrial production ( Pedrozo et al, 2021 ). Therefore, performing simulation calculations before pilot and industrial tests is a viable strategy for effectively reducing trial and error costs.…”