USE OF DETAILED KINETIC MODELS FOR MULTISCALE PROCESS SIMULATIONS OF SULFUR RECOVERY UNITS F. Manenti*, D. Papasidero*, A. Cuoci*, A. Frassoldati*, T. Faravelli*, S. Pierucci*, E. Ranzi*, G. Buzzi-Ferraris* flavio.manenti@polimi.it * Politecnico di Milano, Dipartimento di Chimica, Materiali e Ingegneria Chimica Giulio Natta, Piazza Leonardo da Vinci 32, 20133 Milano, Italy Abstract The modeling of thermal reaction furnaces of sulfur recovery units is a quite cumbersome problem since it involves different modeling scales such as the kinetic/molecular micro-scale, the reactor design meso-scale, and the chemical process macro-scale. The present paper proposes preliminary results of a multiscale approach to model the thermal furnaces and waste heat boiler based on detailed kinetics and reactor network analysis (RNA). The main kinetic mechanisms are discussed and validated using experimental data; industrial data is used to validate the RNA layout. Introduction Process simulation is nowadays supported by many tools and commercial flowsheeting packages involving unit operations, reactors, thermodynamic libraries, and property databases. These tools make possible the simulation of complex processes and overall plants, but they still have certain key-open-issues to be handled to perform accurate simulations and deepen the process understanding. One of the hardest problems is the simulation of non-ideal reactors via detailed kinetic schemes. This lack in the current process simulators is mainly due to: (I) the need of complex and well-established kinetic schemes to characterize the reaction environment; (II) the need to face simulation issues at different scales (kinetic and plant scales); and (III) the need of powerful solvers to handle the resulting largescale stiff nonlinear systems coming from kinetic modeling. The paper investigates the possibility to bridge the gap in process modeling by coupling OpenSMOKE and BzzMath libraries, two freely downloadable tools. OpenSMOKE [1] allows to simulate non-ideal reactors by solving complex networks of ideal elements. It is based on consolidated kinetic schemes [2]. BzzMath library [3, 4] is a numerical library for scientific computing. Specifically, it includes very performing and robust solvers for several numerical areas. It is worth underlining that these tools can be fully integrated in the most widespread commercial packages as discussed elsewhere [5-7]. For its well-known difficulties and renewed academic and industrial interest, the validation case is the thermal furnace of Claus processes, 1
designed to remove sulfur from acid gas streams. Kinetics (microscale) The kinetics of thermal reaction furnace of sulfur recovery units is very complex and not yet completely understood. The kinetics governing the transformations of sulfur compounds has been studied by Mueller et al. [8], who described the main the oxidation mechanisms, and Dagaut et al. [9], who highlighted the inhibition effects of SO 2 on the radical pool. The pyrolysis of hydrogen sulfide, H 2 S, has been defined in detail by different authors [10, 11]. Other authors focused their research on the formation mechanisms of a specific species such as the CS 2 and the COS, but they are not considered in this work for the sake of conciseness, although also the model previsions of these species are in good agreement with the experimental sets that we analyzed. The kinetic mechanisms are collected in an overall kinetic scheme containing 800+ reactions [2], for which the key-reactions only are reported hereinafter as validation. Although in presence of oxygen, the pyrolysis of H 2 S is particularly important in the Claus furnaces, looking forward to its high reactivity at the typical operating conditions and the non-stoichiometric inflow of combustion air. According to the Binoist s reactor and conditions [11] the key steps for the H 2 S pyrolysis are: H 2S = SH + H (1) H 2S = H2 + S (2) Binoist s kinetic parameters are used for (2), whereas Arrhenius parameters 14 k 0 = 210 mol/l/s and E = 66000 cal/mol are proposed for (1). A selection of model previsions related to the reactions above is given in Figure 1 and Figure 2. Conversion (H2S) 60.00% Polimi @ 900 C 50.00% Polimi @ 940 C Polimi @ 970 C 40.00% Polimi @ 1000 C Polimi @ 1050 C Polimi @ 850 C 30.00% Exp. Data @940 C Exp. Data @970 C 20.00% Exp. Data @1000 C Exp. Data @1050 C 10.00% Exp. Data @850 C Exp. Data @900 C 0.00% 0.00E+00 5.00E-01 1.00E+00 1.50E+00 2.00E+00 Residence Time (s) H2 Mole Fraction 3.00E-02 2.50E-02 Polimi @ 850 C Polimi @ 900 C Polimi @ 940 C 2.00E-02 Polimi @ 970 C Polimi @ 1000 C 1.50E-02 Polimi @ 1050 C Exp. Data @ 850 C Exp. Data @ 900 C 1.00E-02 Exp. Data @ 940 C Exp. Data @ 970 C 5.00E-03 Exp. Data @ 1000 C Exp. Data @ 1050 C 0.00E+00 0 0.5 1 1.5 2 Residence Time (s) Figure 1. Pyrolysis (Data: Binoist et al., 2003): H 2 S conversion. Figure 2. Pyrolysis (Binoist et al., 2003): H 2 formation. Under the combustion regime of Claus furnaces, the H 2 S is partially (one third) oxidized to SO 2. The partial oxidation allows to achieve the optimal ratio 2
HS 2 / SO 2 = 2 at the catalytic reactors (Claus converters) to maximize the yield of the overall SRU: 2H S+ SO = 3/ x S + 2H O (3) 2 2 x 2 and thus to maximize the sequestration of elemental sulfur. x accounts for the sulfur equilibrium ( x = 1, 2, 4,6,8 ). More details on the Claus process can be found elsewhere [6, 12]. The oxidation of sulfur compounds can be conveniently described using the analysis of the H 2 S explosion diagram to give SO 2. The sensitivity analysis in correspondence with the slow-oxidation region highlighted the following predominant reactions ordered by relevance: SH + O2 = HSO2 (4) O2 + H2S = HO2 + SH (5) SO + O2 = SO2 + O (6) Conversely, in the explosion region, the reaction (6) is the most important one. The second limit in the explosion diagram defines the passage from the low to the high pressure mechanisms. It is determined by the following competing reactions: SH + O2 = SO + OH (7) SH + O2 = HSO2 (4) The ratio r 7 / r 4 = 1 describes the transition from low to high pressure mechanism. It is possible to evaluate the explosion diagram using the corresponding constants: ( ) PT α [ ][ ] ( ) [ ][ ]( ) ( ) [ ][ ] ( ) 8 [ ][ ] SH O exp E / RT 10 SH O exp 12350 / RT r SH O P / RT exp E / RT 3 10 SH O P / RT 10 r7 7 2 7 2 = = = 4 α4 2 4 2 (8) α E E 10 7 4 7 10 12350 PT ( ) = exp RT exp 64.2T 8 α 4 RT = 310 1.986T (9) The sensitivity analysis performed for the upper limit showed that the reaction (6) in the ignition region is comparable to (4) in the oxidation region. Conversely, in the ignition zone, the following reactions are relevant: SH + O2 = HSO + O (10) 3
H2S+ SO= S2O+ H2 (11) The low pressure limit has poor practical relevance and it is not considered in this work for the sake of conciseness. The oxidation of H 2 S generates SO 2 as major compound. SO 2 is involved in many kinetic mechanisms and, specifically, it plays a key-role in the formation of SO 3 [8]: SO2 + O = SO3 (12) SO2 + OH = HOSO2 (13) HOSO2 + O = SO3 + OH (14) At nominal conditions (lean conditions for combustion air), SO2 can promote or inhibit several mechanisms. For instance, it is a radical pool inhibitor and reduces the CO oxidation rate (Figure 3). Another key-phenomenon is the formation of COS (Figure 4) [13]. 0.006 0.04 Species Mole Fraction 0.005 0.004 0.003 0.002 0.001 Exp. Data, O2 Exp. Data, CO Exp. Data, CO2 Model, O2 Model, CO Model, CO2 COS Mole Fraction (mol%) 0.035 0.03 0.025 0.02 0.015 0.01 0.005 Exp. Data, Reactor 1 Exp. Data, Reactor 2 Model, Reactor 1 Model, Reactor 2 0 0 200 400 600 800 1000 1200 1400 Initial SO2 Mole Fraction (PPM) 0 800 900 1000 1100 1200 T ( C) Figure 3. Inhibition effects (Data: Mueller et al., 2003) Figure 4. COS formation (Data: Karan et al., 2003) Reactor Network Analysis (mesoscale) Thermal reaction furnace and waste heat boiler can be simulated by means of several kinds of reactors in series. This simplified configuration (no computational fluid-dynamics) is useful for control purposes According to the fast ignition of H 2 S with respect to the other species, a perfectly-mixed reactor is adopted to simulate the first portion of the thermal furnace, where the H 2 S is oxidized to SO 2 while the oxygen is available, whereas the remaining species are assumed to be inert as discussed elsewhere [14]. Next, two plug-flow reactors are adopted to simulate the remaining portion of the thermal reaction furnace and the waste heat boiler. The 4
novelty of the approach is also in the use of the detailed kinetic scheme to estimate the recombination effects that take place in this unit [15]. The integration of the detailed kinetic model and reasonable reactor network leads to a comprehensive multiscale simulation for the kernel of sulfur recovery units, consisting of the thermal furnace and the waste heat boiler. The model previsions are compared to the industrial data acquired at Nanjing plant, China (courtesy of Tecnimont-KT S.p.A.). Figure 5 shows that the multiscale (micro and mesoscales) approach allows to properly characterize the behavior of sulfur recovery units. As a result, model previsions are in good agreement with measured outlet compositions (see Figure 6). Only the residual for the CO molar fraction is larger than 1% and it is probably due to fluid-dynamics issues. It is worth underlining that the industrial data available is acquired only at the waste heat boiler outlet, whereas no online measures are physically possible within the furnace: the multi-scale scale model is particularly useful for reliable inference where practical measurements are prevented. 0.04 Species Profile 0.30 Errors in fitting data (After WHB) 0.03 0.03 CO IND. DAT 0.25 Mole Fraction 0.02 0.02 0.01 S2 H2S Simulation Data 0.20 0.15 0.10 H2S SO2 O2 CO CO2 S2 0.01 0.00 O2 0 1 2 3 4 Reactor Length (m) Figure 5. Model previsions and industrial data fitting for a selection of species. 0.05 0.00 0 0.05 0.1 0.15 0.2 0.25 0.3 Exp. Data Figure 6. Residuals; molar fractions for the main species (a zoom for small fractions, on the right). Conclusions The paper proposes the integration of reactor network analysis and detailed kinetic schemes to achieve a multi-scale approach to face the well-known problem of simulating sulfur recovery units. The process scale is not considered for the time being (future developments). Kinetic mechanisms have been validated on extensive literature data. Simulation results of the multi-scale model are in very good agreement with the industrial data provided by Tecnimont-KT. References [1] Cuoci, A., et al., The ignition, combustion and flame structure of carbon monoxide/hydrogen mixtures. Note 2: Fluid dynamics and kinetic aspects of syngas combustion. International Journal of Hydrogen Energy, 2007. 32(15): p. 3486-3500. 5
[2] Ranzi, E., A wide-range kinetic modeling study of oxidation and combustion of transportation fuels and surrogate mixtures. Energy & Fuels, 2006. 20(3): p. 1024-1032. [3] Buzzi-Ferraris, G. and F. Manenti, A Combination of Parallel Computing and Object-Oriented Programming to Improve Optimizer Robustness and Efficiency. Computer Aided Chemical Engineering, 2010. 28: p. 337-342. [4] Buzzi-Ferraris, G. and F. Manenti, BzzMath: Library Overview and Recent Advances in Numerical Methods. Computer Aided Chemical Engineering, 2012. 30(2): p. 1312-1316. [5] Manenti, F., et al., Adaptive Data Reconciliation Coupling C++ and PRO/II and On-line Application by the Field. Computer Aided Chemical Engineering, 2010. 28: p. 373-378. [6] Signor, S., et al., Sulfur Recovery Units: Adaptive Simulation and Model Validation on an Industrial Plant. Industrial & Engineering Chemistry Research, 2010. 49(12): p. 5714-5724. [7] Manenti, F., et al., Process Dynamic Optimization Using ROMeo. Computer Aided Chemical Engineering, 2011. 29: p. 452-456. [8] Mueller, M.A., R.A. Yetter, and F.L. Dryer, Kinetic Modeling of the CO/H2O/O2/NO/SO2 System: Implication for High-Pressure Fall-off in the SO2+O(+M)=SO3(+M) Reaction. International Journal of Chemical Kinetics, 2003. 35: p. 564-575. [9] Dagaut, P., et al., Experimental and Kinetic Modeling Study of the Effect of NO and SO2 on the Oxidation of CO-H2 Mixtures. International Journal of Chemical Kinetics, 1995: p. 563-568. [10] Glassmann, I., Combustion. Academic Press, 3rd Ed., San Diego, 1996: p. 383-398. [11] Binoist, M., et al., Kinetic study of the pyrolysis of H2S. Ind. Eng. Chem. Res., 2003. 42: p. 3943-3951. [12] Manenti, F., M.G. Grottoli, and S. Pierucci, Online Data Reconciliation with Poor-redundancy Systems. Industrial & Engineering Chemistry Research, 2011. 50: 14105-14114. [13] Manenti, F., et al., Reactor network analysis of Claus furnace with detailed kinetics. Computer-Aided Chemical Engineering, 2012. 30(2): p. 1007-1012. [14] Pierucci, S., E. Ranzi, and L. Molinari, Modelling a Claus Reaction Furnace via a Radical Kinetic Scheme. Proceedings of ESCAPE-14, Lisbon, Portugal, 2004: p. 463-468. [15] Manenti, G., et al., Design of SRU thermal reactor and waste heat boiler considering recombination reactions. Procedia Engineering, 2012. 42: p. 414-421. 10.4405/35proci2012.IV1 6