Simulation of partial oxidation process by using flamelet/progress variable approach

Transcription

1 Simulation of partial oxidation proess by using flamelet/progress variable approah M. Vasellari a,, H. Xu a, S. Hartl a, C. Hasse a a Chair of Numerial Thermo-Fluid Dynamis, Tehnishe Universität Bergakademie Freiberg, Freiberg, Germany Abstrat Partial oxidation of a arbonaeous material (natural gas, oal, biomass among others) is widely used for produing gas for power generation or as feedstok in the hemial industry. In partial oxidation proesses two signifiant reations regimes an be observed. In the flame zone, the fast oxidation reations are mainly governed by the mixing of the fuel with the reatants, similarly to onventional lean diffusion flames. In the post-flame zone, the reforming reations take plae, slowly approahing to the hemial equilibrium onditions. In order to desribe both regimes, the flamelet/progress variable (FPV) approah was suessfully applied to model a pilot sale High Pressure Partial Oxidation (HP-POX) reator fired with CH 4. The hemistry was preomputed using premixed freely propagating flames in and stored in a look-up table, after the PDF-averaging of the laminar solution. The laminar solutions were onverted from physial spae to a progress variable spae, whih was defined ombining different hemial speies in order to have a unique mapping. Additionally, the PV should be aurately defined for resolving the slow reations in the post-flame zone. Finally, the PDF look-up table was oupled to the CFD solution of the turbulent flow in the HP-POX reator. The numerial results show that the FPV allows to desribe both the fast oxidation and the slow reforming reations, giving a reasonable agreement with the experiments. In addition to the preise preditive apabilities, the omputational ost by using the FPV approah is signifiantly redued due to the effiient lookup of the preomputed flamelet solutions. Introdution Partial oxidation is a proess in whih a solid, liquid or gaseous hydroarbon feed in a fuel rih mixture is onverted into a synthesis gas (syngas), primarily omposed of hydrogen and arbon monoxide. Suh mixtures represent the intermediary step from hydroarbons to bulk hemials (methanol, DME, ammonia, aeti aid, oxo-alohols, isoyanates) and syntheti fuels (syntheti Diesel) []. CFD simulations of a High pressure Partial Oxidation (HP-POX) pilot-sale reator were performed using the Eddy Dissipation Conept (EDC) [2] with a detailed mehanism redued from GRI-meh [3]. EDC is very omputationally expensive, espeially in ombination with the large kineti mehanism required for desribing the slow reforming proess. To redue the omputing effort for simulating the turbulene-hemistry interation redution tehniques are usually applied. The hemistry is separated from the solution of the turbulent flow and modeled using simplified redued approah, suh as homogeneous reators, D diffusion and premixed flames, generally referred as flamelets. Results of the redued hemistry models are pre-alulated and stored in a flamelet lookup table (FLUT). The tables are ommonly parametrized using global salar variables, suh as mixture fration Z and progress variable Y, et.. FLUTs are oupled to the CFD solutions through the same input variables, whih are solved together with the turbulent flow. The turbulene-hemistry interation is aptured with the PDF integration based on the mean/filtered values and its variane of mixture fration (and progress variable). The Corresponding author: Mihele.Vasellari@vt.tu-freiberg.de Proeedings of the European Combustion Meeting 25 steady laminar diffusion flamelet (SLDF) approah [4] was used for modeling the HP-POX reator [5]. However, different reation regimes an be identified in the HP-POX. On one hand, the flame zone is generally haraterized by fast oxidation reations, whih is orretly desribed by the SLDF approah. On the other hand, the reforming zone is dominated by slow reations onverting the ombustion produts with exess fuel into CO and H 2. This zone mainly ours in premixed regime and with larger time- and length-sale than the oxidation zone, and it annot be properly desribed by the SLDF [5, 6]. The flamelet/progress variable (FPV) [7] model is a suitable hoie to apture slower proesses as demonstrated for loal extintion, NO x emission and radiation [8, 9]. The FPV model was derived from the SLDF approah, as well as the flamelet generated manifolds (FGM) approah []. The FPV model use the idea of the progress variable as the parameter desribing the progress in the reating mixture. In this work the FPV approah is used for desribing the turbulene-hemistry interation in the HP-POX reator [2]. The hemistry is modeled by means of freely propagating premixed D flames and stored in premixed flamelet look-up tables (P-FLUT). The use of a P-FLUT is justified by the fat that the large part of the HP-POX reator is dominated by the slow reforming reations, whih mainly our in premixed regime. Numerial models The turbulent reating flow in the HP-POX reator is simulated using the 2D axisymmetri version of the ommerial CFD solver Ansys-Fluent [], using the pressureveloity oupled algorithm. Convetive fluxes in all transport equations are disretized with a seond-order upwind sheme and the pressure gradients are disretized with

2 Figure : Shemati view of HP-POX reator. a seond-order aurate sheme. Turbulene is modeled using the realizable k ε approah [2], and radiation with the P- model [3]. The radiating properties of the gas were modeled assuming a gray-band model, based on the Weighted Sum of Gray Gases (WSGG) model [4]. The FPV approah is implemented in Ansys-Fluent using the UDF apabilities of the software and oupled to the pre-alulated FLUT. The look-up and interpolation was performed with a multi-dimensional interpolation proedure and a subsidiary memory abstration layer [5]. Governing equation for the turbulent reating flow In the FPV approah the transport equations for the mean ( Z) and the variane ( Z 2 ) of the mixture fration and for the mean progress variable (Ỹ) are solved in addition to the transport equations for the momentum, pressure, total enthalpy, turbulent kineti energy and its dissipation: (ρu Z) = [ρ(d + D t ) Z] () (ρu Z 2 ) = (ρ(d + D t ) Z 2 ) + C g ν t ρ ( Z) 2 C d ρ ɛ k Z 2 (2) (ρuỹ) = [ρ(d + D t ) Ỹ] + ω, (3) Where u is the veloity vetor, D the moleular diffusivity and D t the turbulent diffusivity. For all the transport equations the turbulent diffusivity was alulated with the turbulent Shmidt-number being S t = ν t D t =.7, where ν t is the turbulent visosity. For all moleular diffusion oeffiients, a unity Lewis-number approah λ Le = ρ pd was assumed. The onstants terms in Eq. 2 are C g = 2.86 and C d = 2. [6]. The density ρ, the progress variable soure term ω Y, the temperature T and the speies mass frations Ỹi are looked-up from the FLUT table using Z, Z 2 and Ỹ as input parameters. The adiabati temperature and density from the FLUT are orreted aounting for the enthalpy defet due to the radiative heat flux inside the reator. Chemistry The Cantera software pakage was used to model onedimensional adiabati steady laminar freely propagating premixed flames in order to generate the FLUT required by the CFD simulation. The transport properties were 2 evaluated assuming Le =, similarly to the CFD solution. Furthermore, an adaptive grid refinement was used with an average of over grid points on a domain up to 5 m to ensure a hemial equilibrium state in the burned gas, espeially for fuel rih mixture fration, where the slow reforming reations requires a very long domain for ahieving hemial equilibrium. The hemial mehanism developed at the Deutshes Zentrum für Luft- und Raumfahrt (DLR) [7] is used for solving the D flames. The DLR mehanism was originally developed to aount for the PAH gas phase growth in methane, ethylene and ethane flames. Modifiations to the gas phase hemial mehanism are proposed in order to inrease the quality of the predition of soot preursors. These hanges in the mehanism were also tested on premixed methane, ethane and ethylene flames with oxygen to ensure that the hanges in the mehanism do not orrupt the appliation to C 2 fuels. The proposed modifiations improve the agreement of the premixed laminar flame simulations when ompared to the experimental results. The used version of the DLR mehanism onsists of 43 speies and 39 reations. For eah mixture fration the premix flame is onverted from physial spae to progress variable Y spae. The progress variable is defined as linear ombination of speies onentrations: n Y = α i Y i. (4) i= A suitable hoie of the progress variable is ruial. Attention has been paid to this topi in [8, 9, 2, 2]. Some requirements for the progress variable an be defined aording to [9]: (i) the transport equation is onveniently solved in a ombustion simulation. (ii) The parameter setup haraterize uniquely eah point in the thermo-hemial state spae, hene it has to be stritly monotonially dereasing or inreasing. (iii) Proess ontrols quantity and projetion of the trend of the ombustion (flame region, slow onversion reations). (iv) Appropriate mapping with respet to the speies and temperature is required. For omplex flame strutures and detailed hemistry identifying a suitable progress variable an beome more ompliated. Furthermore, turbulent effet are aounted by a PDFintegration of the laminar flame database, whih onsiders the mean mixture fration and its variane to apture the

3 Oxidizer Gasifiation Agent Fuel ṁ, kg/s T, K CH 4 O H 2 O.55 N 2.8 Z.. Y.55 Table : Input onditions of the HP-POX reator. Conentrations are given as mass frations. (a) T, K T, K 24 T 2 ω ω, kg/(s-m 3 )/ 5 4 Y i.6.4 CH 4 O 2 CO CO 2.2 (b) ω, kg/(m 3 s) Figure 2: Temperature and progress variable soure term in mixture fration/progress variable spae from P-FLUT. Figure 3: Temperature, progress variable soure term and mass fration of the main speies as a funtion of the normalized progress variable at Z =.388 from P-FLUT. turbulene-hemistry interations. The PDF for the mixture fration is a presumed β-pdf. For onveniene, the mixture fration and progress variable were assumed to be statistially independent and the progress variable PDF was assumed to be a δ-pdf. After the integration, the progress variable is normalized with = Y Y,min Y,max Y,min and the P-FLUT was onstruted with Z, Z 2 and C being the look-up parameters. High Pressure Partial Oxidation (HP-POX) pilotsale reator Detailed desription of the HP-POX pilot-sale reators was reported by Rehm et al. [2]. Figure reports the shemati view of the reator. Three input streams are onsidered in the reator, as illustrated from the enlarged view of the burner: the oxidizer stream, omposed by a mixture of O 2 and H 2 O, enters from the entral inlet, 3 the gasifiation stream, omposed by water enters from the intermediate annular setion, and the fuel, omposed by CH 4 from the external annular setion. In order to simplify the FPV approah only two streams are onsidered for the alulation: the fuel stream with Z = and the oxidizer stream with Z =, whih orresponds to the mixing of the oxidizer and the gasifiation streams. The mass flow rates, omposition and temperature of the input streams are reported in Tab.. Reator walls are assumed adiabati. For the simulations, a 2D axisymmetri mesh with 4 hexahedral ells was used after grid independene ould be ensured. Results Analysis of the P-FLUT The P-FLUT was generated using freely propagating premix flames as previously desribed. The progress variable is defined as follows: Y = 2Y H2 + Y H2 O +.6Y CO2 +

4 (d) (a) T, K (b) Z () Y Figure 4: Temperature (a), mixture fration (b), progress variable () and normalized progress variable (d) distributions on the HP-POX axial setion obtained from the CFD FPV simulation. The blak lines show the stoihiometri onditions (Zst =.92) YCO. Y is hosen in order to satisfy the requirements desribed above, and partiular attention was given for aurately resolving the partial oxidation proess lose to the outlet mixture fration (Zout =.388), whih is the main goal of the work. for the outlet mixture fration. The soure term ω is about three orders of magnitude lower than the ones at the stoihiometri onditions (Fig. 2). Consequently the time and the length sales of the reforming zone are muh respet to the flame zone. The temperature inreases almost linearly with the progress variable up to.67, whih orresponds to the point where the O2 is almost ompletely onsumed. One O2 is fully onsumed, the endothermi reforming reations proeed extremely slowly, reduing the temperature and almost ompletely onsuming the remaining CH4. Additionally CO inreases during the reforming. The temperature and the progress variable soure term obtained from the laminar D premix flames are shown in the Z-Y spae in Fig. 2. The lower limit in the plots represent the minimum values of Y for a given mixture Z, orresponding to =. This ondition represents the pure mixing between the reatants without reations. The non-normalized progress variable is set to zero for the fuel inlet (Z = ), while is set to.55 for the oxidizer inlet (Z = ), orresponding to the mass fration of H2 O in the oxidizer. The upper limit represents the maximum values of Y, whih orrespond to the equilibrium ondition reahed by the premix manifold for every mixture. Here the normalized progress variable assume a value of one. The maximum value of Y =.39 is obtained lose to the outlet mixture (Zout =.388), optimizing the resolution of the progress variable in the reforming zone of the HPPOX reator. The solutions of the D premixed flames are obtained for several mixture frations going.7 to.92. Outside of this range the results of the extreme solutions are extrapolated to the minimum and maximum values of the mixture fration spae. Analysis of the CFD results The results of the HP-POX reator simulation using the FPV approah with a P-FLUT are reported in this setion. Figure 4 shows the ontour plots of the mean temperature, mixture fration, non-normalized and normalized progress variable on the axial setion of the HP-POX reator. The stoihiometri onditions are also reported in the plots (Zst =.92). The normalized progress variable (Fig. 4d) is also reported in addition to the non-normalized Y for highlighting the deviation from the equilibrium onditions inside the reators. A sale between.9 to is used for for underlining the onversion in the reforming zone. The largest gradients an be observed in the oxidation region lose to the reator burner, while the remaining reforming part is haraterized by extremely low variations of the flow properties. A zoomed view of the oxidation region is therefore reported for every salar for highlighting plot shows that the gradients in the oxidation zone. The Z mixing between the fuel and the oxidizer ours in about / of the total length of the reator. However the flame is muh longer than the size of the burner, due to the high veloities of the input streams. The external fuel stream is also mixed to the reirulating syngas. The extremely high reation rates, related to the oxyfuel onditions in the Very high temperatures (Fig. 2a) above 3 K are ahieved around the stoihiometri mixture fration (Zst =.92), orresponding to extremely high values of the progress variable soure term (Fig. 2b). Suh high temperatures and soure terms mainly depend on the very high O2 ontent of the oxidizer streams (84 wt% see Tab. ). It means that equilibrium onditions are quikly reahed for values of mixture frations lose to the stoihiometri onditions. Riher mixtures are haraterized by slower reation rates and temperatures. For this reason T, ω and the main Yi from the P-FLUT are also reported in Fig. 3 4

5 reator, quikly push the reations towards the hemial equilibrium, as it may be observed from Figs. 4()-(d). Externally to the mixing zone, whih is bounded by the stoihiometri onditions, the mixture fration quikly reahes the outlet ondition ( Z =.388), behaving like a premixed reforming reator. In the reforming zone the temperature and the other salars are haraterized by an uniform distributions. The Y distribution (Fig. 4) shows values lose to.8 along the stoihiometri onditions, whih orresponds to the maximum Y for Z st. In the reforming zone the progress variable shows an almost uniform distribution, due to the slow advanement of the reforming reations. For learly highlighting the slow advanement of the reforming reation, the normalized progress variable is also shown (Fig. 4d). In fat, in the oxidation zone, quikly reahes the equilibrium onditions ( = ), as previously disussed. Downstream of stoihiometri onditions, the normalized progress variable dereases from one to values between.94 to.98. The derease of the normalized progress variable depends on the mixing with riher mixtures with <. These mixtures are haraterized by higher values of the maximum non-normalized progress variable, as it an be observed from Fig. 2 moving from Z st to Z =.388. The numerial results of the CFD simulation of the HP-POX are also reported along the enterline in Fig. 5. As already observed from the ontour plots in Fig. 4, the mixture fration Z at first gradually inreases, then rapidly inreases at y/l =. rossing the flame, and remains nearly onstant in the reforming zone for y/l >. Similar behavior an be observed for the non-normalized progress variable Y, whih reahes the maximum values Y =.22 approximately at y/l =. for Z =.365. After this point Y slightly dereases and than inreases again. The plot also reports the maximum progress variable Y,max for the given mixture fration. For y/l <. the progress variable orresponds to its maximum value, and only in the reforming zone beomes lower than Y,max. The differene between Y and Y,max an be more learly observed looking at the normalized progress variable plot. In fat, rapidly tend to in the oxidation zone, beause of the high oxidation rates, while in the reforming zone at first dereases to.967, beause of the mixing. Than it begins slowly to inrease until the exit of the reator without reahing the equilibrium onditions. The main hemial speies onentrations are also reported as mass frations. In the oxidation region, O 2 is ompletely onsumed as soon as the stoihiometri onditions are reahed at y/l =.. CO follows the opposite trend, quikly inreasing in the flame and then slowly in the reforming zone. CO 2 reahes its maximum for the stoihiometri onditions and than it slowly dereases in the reforming zone. Small frations of CH 4 an be observed after the stoihiometri onditions from the mixing of the burned produts with the reirulated syngas. Finally, Fig. 6 shows the omparison between the experimental outlet omposition in mass fration [2] and the numerial results for the HP-POX reator. The CFD results 5 Z, Y T, K Y i Y Y,maxZ y/l O 2 CH 4 CO CO 2 Figure 5: Mixture fration, progress variable, temperature and speies mass frations along the enterline of the HP-POX reator. show generally a good agreement with differenes lower than 2.8% by weight. CO, whih is the most abundant speies, is orretly predited with a deviation of.3%. The largest deviations exist for CO 2 and CH 4. In partiular the numerial simulation predits.6% of CH 4, while only.5% was measured. The omparison of the progress variable Y shows a very good agreement with a differene of.%, whih onfirms the validity of the approah used for desribing the slow reforming reation rates. Conlusion A pilot-sale high pressure partial oxidation reator (HP-POX) was numerially investigated using the flamelet progress variable (FPV) approah oupled with a premix laminar flamelet look-up table (P-FLUT). The HP-POX is haraterized by an oxidation zone, whih is mainly governed by the diffusion of the reatants, and by a reforming zone, where the slow hemial reations take plae in premixed regime. For this reason, the FPV approah, whih was originally developed for partially premixed flames, allows to desribe the different reation regimes existing in the reator. Sine the main fous of this work is too aurately desribe the slow reforming proess, the CFD solution of the reator is obtained oupling the FPV approah to a look-up FLUT table generated using D laminar freely propagating premix flames. The hoie of

6 Y i th Exp CH 4 CO CO 2 H 2 H 2 O Y Figure 6: Comparison between the measured and alulated syngas outlet omposition for the HP-POX reator reported as mass fration. the progress variable is one the most deliate hallenge for using the FPV in partial oxidation flames. In fat, the progress variable should allow to desribe the reforming reations with a suffiient resolution, beside to be always stritly monotoni for all the mixture frations. The analysis of the P-FLUT shows that hemial reations in proximity of the stoihiometri onditions are extremely fast, beause the fuel reats with almost pure oxygen. Therefore the oxidation reations our at extremely smaller time and length sales ompared to the reforming. In fat, the CFD analysis of the HP-POX reator using the FPV approah in ombination with the premix look-up table shows that a large reator volume is required for ompleting the reforming proess. The oxidation zone takes up only a minor part of the reator, where the mixing of the reatants our. The FPV approah in ombination with a look-up table generated using D freely propagating laminar flames gives an exellent agreement with the measurements of the outlet syngas omposition, showing differenes of the mass frations of the main speies below 2.7% by weight. In partiular the outlet progress variable shows an exellent agreement with the experimental one, demonstrating that the partial oxidation proesses an be aurately modelled using the flamelet/progress variable. Aknowledgments This researh has been funded in the framework of Virtuhon by the Federal Ministry of Eduation and Researh of Germany (projet number 3Z2FN) and by the Saxon Ministry of Siene and Fine Arts (projet number /). Referenes [] T. Christensen, I. Primdahl, Hydroarbon Proessing 73 (994) [2] M. Rehm, P. Seifert, B. Meyer, Computers & Chemial Engineering 33 (29) Gri-meh - an optimized detailed hemial reation mehanism for methane ombustion, 995. [4] N. Peters, Proeedings of the Combustion Institute 2 (986) [5] S. N. P. Vegendla, D. Messig, S. Weise, C. Hasse, Energy Fuels 25 (2) [6] F. Hunger, B. Stelzner, D. Trimis, C. Hasse, Flow, Turbulene and Combustion 9 (23) [7] C. Piere, P. Moin, Journal of Fluid Mehanis 54 (24) [8] M. Ihme, C. M. Cha, H. Pitsh, Proeedings of the Combustion Institute 3 (25) [9] M. Ihme, H. Pitsh, Phys. Fluids 2 (28). [] J. V. Oijen, L. D. Goey, Combustion Siene and Tehnology 6 (2) [] ANSYS, In., ANSYS-FLUENT V 4.5 Commerially available CFD software pakage based on the Finite Volume method. Southpointe, 275 Tehnology Drive, Canonsburg, PA 537, U.S.A., [2] T. Shih, W. Liou, A. Shabbir, Z. Yang, J. Zhu, Computers and Fluids 24 (995) [3] P. Cheng, AIAA Journal 2 (964) [4] T. F. Smith, Z. F. Shen, J. Friedman, The ASME Journal of Heat Transfer 4 (982) [5] S. Weise, D. Messig, B. Meyer, C. Hasse, Combustion Theory and Modelling 7 (23) [6] W. P. Jones, J. H. Whitelaw, Combustion and Flame 48 (982) 26. [7] V. Chernov, M. J. Thomson, S. B. Dworkin, N. A. Slavinskaya, U. Riedel, Combustion and Flame 6 (24) [8] A. Najafi-Yazdi, B. Cuenot, L. Mongeau, Combustion and Flame 59 (22) [9] M. Ihme, L. Shunn, J. Zhang, Journal of Computational Physis 23 (22) [2] Y.-S. Niu, L. Vervish, P. D. Tao, Combustion and Flame 6 (23) [2] U. Prüfert, S. Hartl, F. Hunger, D. Messig, M. Eiermann, C. Hasse, Flow Turbulene and Combustion (25). Doi:.7/s [3] G. Smith, D. Golden, M. Frenklah, N. Moriarty, B. Eiteneer, M. Goldenberg, C. Bowman, R. Hanson, S. Song, W. Gardiner, V. Lissianski, Z. Qin, 6