A reactive distillation process for deep hydrodesulfurization of diesel: Multiplicity and operation aspects

Abstract

A systematic study of the operating conditions and parameter sensibility under which multiple steady states may occur in a reactive distillation column (RDC) for diesel deep hydrodesulfurization is presented. The multiplicity analysis is performed through bifurcation diagrams for two feed case scenarios. The main variables that affect the steady state behavior are the reflux ratio, the hydrocarbon feed flowrate, the reboiler heat duty, and the catalyst load (liquid holdup). For the first case only a single steady state was found, while in the second case input and output multiplicities were determined. It is shown that the introduction in the feed of the recalcitrant 4,6-DMDBT, plays a crucial role in the RDC design and operation. A set of values of the bifurcation parameters for an appropriate operation of the RDC are recommended to avoid the multiplicity region and to reach the required output targets.

Introduction

Conventional hydrotreating is a commercially proven refining process that passes a mixture of heated feedstock and hydrogen (H₂) through a catalytic reactor to remove sulfur and other undesirable impurities. A review of the technologies for producing ultra-low sulfur diesel (ULSD) fuel reveals that current technologies can be modified to produce diesel with less than 10 parts per million (ppm) of sulfur. Nevertheless, only a small number of refineries currently produce diesel with sulfur in the 10 ppm range on a limited basis. The existence of the required technology does not ensure that all refineries will have that technology in place and in time to meet the new ULSD standards because these plants are characterized by a wide range of size, complexity, and quality of crude oil inputs.

It is generally believed that a two-stage deep hydrodesulfurization (HDS) process will be required by most, if not all refiners, to achieve a diesel product with less than 10 ppm of sulfur. A design consistent with this technological approach has been proposed by Haldor Topsøe (DOE/EIA Report, 2001), which includes a first stage that reduces the sulfur content to around 250 ppm or lower, and a second stage that completes the reduction to less than 10 ppm. Knudsen, Cooper, and Topsøe (1999) suggested that to deep desulfurize cracked stocks, a higher operating pressure of the reactor is necessary. Pressure requirements would depend on the quality of the crude oil and the setup of the individual refinery; thus, the level of pressure required for deep desulfurization is a key uncertainty in assessing the cost and availability of the technology. Based on the philosophy of previous works and designs (Sakanishi et al., 1991, Sakanishi et al., 1992, Takatsuka et al., 1997), it could be considered a two-stage deep desulfurization process where the first stage is a currently conventional HDS process that reduces the sulfur content up to 500 ppm, and a hypothetical second stage being a reactive distillation column (RDC), which should remove the sulfur from diesel to the desired specification. This RDC would operate at lower pressures and, hence, it will reduce H₂ consumption, and energy requirements.

Reactive distillation is considered as a highly promising process because it combines the requirements of in situ separation with reaction. This integration brings several advantages, i.e., energy and capital savings, increased reactant conversion, enhanced product selectivity, and improved heat integration. Nevertheless, the presence of chemical reactions in a distillation column leads to difficulties in the modeling (Taylor & Krishna, 2000). The increasing interest in reactive distillation has been accompanied by the development of various simulation algorithms in order to study the operation and control of this process (Al-Arfaj and Luyben, 2002, Hung et al., 2006, Kienle and Marquardt, 2003, Kumar and Kaistha, 2008, Monroy-Loperena et al., 2000, Wang et al., 2003).

Up to now, only few papers have addressed the application of reactive distillation to the diesel deep HDS. Taylor and Krishna (2000) discussed the possibility of applying the reactive distillation concepts to crude oil fractions HDS, while Krishna (2002) showed how this technology could be used. An analysis of the operating conditions to obtain ULSD in a conventional HDS process (Knudsen et al., 1999, van Hasselt et al., 1999) suggests that reactive distillation could be an interesting technological alternative for diesel deep HDS. In a reactive distillation process, the countercurrent flow is the natural operation mode and the internal flowrates requirement can be obtained through the catalyst packing arrangement, regulating the reflux and/or the boilup ratio, and placing properly the sulfured hydrocarbon feed. Viveros-García, Ochoa-Tapia, Lobo-Oehmichen, de los Reyes-Heredia, and Pérez-Cisneros (2005) showed a comparison between deep HDS in a conventional countercurrent trickle-bed reactor and the operational and design alternatives offered by a reactive distillation process. They noted that all the operation requirements for a deep HDS in the reactor could be fulfilled by a reactive distillation operation with an appropriate process design.

Design, operability, and control of a RDC become more difficult due to the interactions between chemical reaction and separation. Such complex interactions, primarily between kinetic expressions and thermodynamic models, lead to a highly nonlinear behavior of the RDC indicating the possible existence of multiple steady states (MSS). Different authors have identified MSS (input and output multiplicities) in conventional, azeotropic, and reactive distillation systems (Baur et al., 2003, Ciric and Miao, 1994, Gani and Jørgensen, 1994, Jacobsen and Skogestad, 1991, Kannan et al., 2005, Kienle and Marquardt, 2003, Mohl et al., 1997, Mohl et al., 1999, Mohl et al., 2001, Müller and Marquardt, 1997, Singh et al., 2005, Wang et al., 2008, Yang et al., 2006). Some of them concluded that MSS can arise in binary distillation if the flowrates are given on a mass basis instead of a molar basis. Gani and Jørgensen (1994) reported MSS in heterogeneous azeotropic distillation columns through the study and simulation of the separation of ethanol–water–benzene ternary system. They pointed out that the movement of the internal composition front could be the cause of the possible connection between multiplicity and operability. They realized that the system reached a different steady state as fronts corresponding to temperature and composition profiles move up or down depending on small positive or negative pulse disturbances in the reboiler heat duty. Müller and Marquardt (1997) verified the experimental existence of MSS for first time in a heterogeneous azeotropic distillation by rigorous simulations and bifurcation diagrams. Jacobsen and Skogestad (1994) studied the stability of conventional (non-reactive) distillation columns and showed that columns displaying MSS had at least one solution that corresponded to an unstable point and that instability was more likely with large internal flowrates. The RADFRAC module (from ASPEN PLUS process simulator) has been used by Jacobs and Krishna (1993) to study MSS in a RDC for MTBE synthesis. Their results showed high and low conversion steady states when methanol is fed to stages 10 or 11 using the Jacobs–Krishna column configuration. Physical explanation for the occurrence of MSS in the MTBE process was provided by Hauan et al., 1995, Hauan et al., 1997. Mohl et al. (1999) employed a pilot-scale column to produce MTBE and TAME. MSS were found experimentally when the column was used to produce TAME, but not in the MTBE process. Baur et al. (2003) presented a bifurcation analysis for the synthesis of TAME in a RDC with two different methods for describing the reaction kinetics: pseudo-homogeneous models and heterogeneous models. Both reaction models showed the possibility of MSS. Therefore, it is evident that the analysis of existence of a single steady state (SSS) or MSS should give insights into the reactive distillation process, help to avoid unsafe operating conditions, and further facilitate subsequent studies such as control, monitoring, data reconciliation, parameter estimation, and optimization of existing reactive distillation process.

In academia and industry the nonlinear analysis of chemical engineering models has great significance (Krasnyk, Ginkel, Mangold, & Kienle, 2007). In particular, the steady state and dynamic simulations of chemical processes can be done through two computational approaches: oriented-equation models and modular models. Flowsheet simulators for chemical process, based on modular models (such as ASPEN PLUS, HYSYS, PRO/II, and CHEMCAD), have the main advantage of including: property databases, thermodynamic models, process unit models, numerical methods, etc. However, a drawback of them is that they do not provide algorithms for nonlinear analysis explicitly (i.e., bifurcation algorithms and stability analysis). While working with equation-based models, it is possible to use continuation algorithms included in packages such as AUTO, CONT, HOMPACK, PITCON, BIFPACK, DIVA (see Mangold, Kienle, Gilles, & Mohl, 2000 and references in) or ProMoT (Krasnyk et al., 2007). Specifically DIVA has been used to study MSS in a RDC of MTBE and TAME (Katariya et al., 2008, Mangold et al., 2000). Some reported benefits of DIVA is that it includes a model library with models of standard operation units and new models are written in symbolic form, also it comes with an inbuilt package for continuation and stability analysis for DAEs systems. On the other hand, notwithstanding the modular simulators do not allow detailed nonlinear analysis, the aim of several works has been to develop procedures for the construction of bifurcation diagrams in modular simulators (Kannan et al., 2005, Vadapalli and Seader, 2001, Yang et al., 2006). Vadapalli and Seader (2001) proposed two continuation algorithms that can be added into ASPEN PLUS to calculate steady state branches in regions of multiplicity. The natural arclength algorithm for an adiabatic continuous stirred tank reactor problem and pseudo-arclength algorithm for a homogeneous azeotropic distillation problem were successful in detecting both turning points and all three steady state branches. However, the limitations were the impossibility of determining whether the steady states were stable or unstable, and of detecting exact bifurcation points because the Jacobian matrix generated by ASPEN PLUS is not accessible by the user. Kannan et al. (2005) identified MSS in homogeneous azeotropic distillation using the HYSYS process simulator in steady state mode, while the stability of the different branches was determined by dynamic simulations also in HYSYS. Nevertheless, most of the reported works have been for conventional (non-reactive) distillation columns or chemical reactors, but for reactive distillation columns (RDCs) few works have considered the construction of bifurcation diagrams using modular simulators. Yang et al. (2006) found input and output multiplicities in a RDC for synthesis of ethylene glycol, defining a procedure in ASPEN PLUS with the method of sensitivity analysis, but without establishing the stability of the branches.

Thus, in this paper we present a systematic study of the operating conditions and parameter sensibility under which MSS may occur, and to assess their influence on a RDC for the deep HDS of diesel. A procedure for multiplicity analysis in ASPEN PLUS based on bifurcation diagrams is presented. The multiplicity analysis is performed for two feed case scenarios, showing its implication on the design and operation of the RDC. Specifically, we study the effect of the operating conditions and parameter sensibility over the main variables to monitor or control hereinafter: the recalcitrant organo-sulfur compounds conversion and the product purity.