Development of Simplified Cracker Yield Model Correlation based on SPYRO~reg; Rigorous Simulator for Various Operational Optimization Applications

摘要

Ethylene producers around the world face increasingly stiff competition in the market place. One of the keys to its competitiveness is the ability of a producer to evaluate and secure the most desirable feedstock and to process it at optimal operating conditions. To achieve this goal, advanced optimization planning and scheduling applications that represent the real business problem are required. Optience SCMart has been selected by COPESUL as the application platform for building nonlinear and mixed integer models. Without the limitation on the optimization technology for solving a realistic model, the key in matching the real operation data is in estimating the yield of the crackers. SPYRO~®, a rigorous cracker yield model from Technip, has been selected as the basis for developing a simplified low order nonlinear polynomial correlation model that is suitable to be embedded in the SCMart planning and scheduling model. SPYRO~® is the industry standard for pyrolysis yield prediction. Its fundamental kinetic scheme of more than 3000 total reactions, representing the chain reaction mechanism of 128 molecular species and 20 radical species is well known for its accuracy in predicting pyrolysis yields for feedstock types, ranging from gas to naphthas/condensates and heavy gasoils. There have been reports of utilizing SPYRO~® in real-time optimization (RTO) and planning applications. These applications are typically either a rigorous equation based RTO or simpler LP based planning. There are basically two approaches that have been reported in the literature: (1) a direct call to the SPYRO~® simulator as a 'black box' during the optimization for a function evaluation and first order numerical derivative, needed for the nonlinear optimization algorithm in the RTO application or (2) a simple 'look-up' table or first order Taylor's series expansion that has been generated a priori for the LP based optimization planning application. In this paper, we propose a low order nonlinear polynomial correlation yield model based on SPYRO~® results. We discuss our experience in obtaining a good model over the desirable range of operating conditions and feedstock types considered in the business operation policy. The model was developed using Optience's SCMart Estimator. In addition to its main function as a nonlinear parameter estimator for the optimization software, its database is designed to manage user feedstock data and SPYRO~® yield results in an efficient way. SCMart Estimator is a module in Optience's SCMart Suite for Planning and Scheduling software, where its optimization solution approach is based on a mathematical programming framework that can solve a wide range of model types, from simple LP to more complex MINLP models.