Operation optimization and decoking scheduling of industrial naphtha cracking furnaces using a simplified furnace model

Abstract

A naphtha cracking furnace produces various petrochemical products ranging from ethylene to pitch. The simulation and optimization of the cracking furnace in NCC have attracted attention of many plant engineers and researchers because the cracking furnace is the most energy consuming unit in NCC. It consumes about 55-65% of the total energy cost. However, due to the complexity of cracking reactions, many of their studies had been limited to the ethane or propane cracking furnaces. But many NCC companies in Asia and Europe use naphtha as their main feed material, the optimization of naphtha cracking furnaces is essential to them. Although there have been some commercial furnace simulators, they have been used only as yield predictors because they give no guideline for the optimum operation. In this study, we developed a simplified furnace model using neural networks. The training and validation data for this model were obtained over a wide range of operating bounds by a furnace simulator, CRACKER.With this furnace model, we formulated an MINLP problem maximizing the total profit in a given time horizon. Tube skin temperature and pressure drop inside the pyrolysis coil are constrained by the plant operation limits. This problem was solved for one furnace operation by GAMS. The optimal decoking interval and inlet naphtha feed flowrate trajectory were obtained by solving the optimization problem. Continuous operation of the furnace leads to coke formation on the inner surface of the pyrolysis coils, the amount of which increases with on stream time. The coke inside the coils decreases productivity of the furnace and therefore its operation is periodically stopped for decoking in order to restore the productivity. A furnace system has many furnaces in parallel, which share the inlet naphtha feed flow to the system. This leads to an optimization problem where the decoking schedule for a set of parallel furnaces must be decided simultaneously with the distribution of...