Bridges in modelling and simulation of steam cracking: from fossil to renewable feedstock and from molecule to furnace

Size: px

Start display at page:

Download "Bridges in modelling and simulation of steam cracking: from fossil to renewable feedstock and from molecule to furnace"

Leon Stafford
5 years ago
Views:

1 Bridges in modelling and simulation of steam cracking: from fossil to renewable feedstock and from molecule to furnace Kevin M. Van Geem andguy B. Marin Laboratory for Chemical Technology, Ghent University 66th Canadian Chemical Engineering Conference Sustainability & Prosperity October (2016) Québec City (Canada) 1

2 Frommolecule toindustrialplant Process level Molecular level 2

3 Steamcracking: fromfossiltorenewables Crude oil Steam cracking Consumer goods from chemical industry Naphtha Ethene Light gasoil crude oil Natural gas Propene Natural gas liquids Butadiene Bio-based feeds Gascondensates Hydrodeoxygenated FAME Aromatics atozforex.com; pnnl.org; districtenergy.org; scade.fr; schmidt-clemens.de; Linde Group 3

4 Steam cracking: Ghent University history 4

5 Steamcracking: hot section Preheat feed and other utility streams 5% Saturated steam 700K Feed Dilution steam 50% C FPH ECO SSH HTC D 800K 1450 K BFW Steam A B K 45% Rapidly quenching of reactor effluent A: Radiation section B: TLE C: Steam drum D: Convection section Endothermic process K 5

6 Analytical Goal techniques Reactionandreactor model Reactor geometry Operating conditions Feedstock properties Simulation model Product specs Modeling Microkinetic model Reactor model Detailed GC GC feedstock composition Fundamental reactor model Detailed product composition 6 Ristic, N.D. et al.,journal of Visualized Experiments, Issue 114Toraman, H.E. et al., Journal of Chromatography A, 1460, , 2016

7 Outline Introduction Feedstock Kinetics Reactor Process Conclusions 7

8 Outline Introduction Feedstock: characterization Kinetics Reactor Process Conclusions 8

3D view 66th Canadian Chemical Engineering

16-19 (2016) Québec City (Canada) On-line

,journal of Chromatography A, 1218, 3217-3223,

9 3D view 66th Canadian Chemical Engineering Conference Sustainability & Prosperity October (2016) Québec City (Canada) On-line GCxGC Pyl, S.P. et al.,journal of Chromatography A, 1218, , 2011 GCGC chromatogram: 2 parts Conventional 1D part C 4- Comprehensive 2D part C 5+ State-Key Laboratory of Chemical Engineering@ECUST September 22,

10 Analytical techniques SIMCO: ANN or Shannon entropy Goal Reactor geometry Operating conditions Standard methods Feedstock properties Simulation model Product specs Feedstock reconstruction SIMCO Microkinetic model Reactor model GC GC Detailed feedstock composition Fundamental reactor model Detailed product composition 10

11 Maximization of Shannon Entropy Feedstock properties Average molecular weight Elemental composition Specific density Global PINA analysis Boiling point data (e.g. D2887 simdist) Aromatic Sulfur Feedstock Shannon reconstruction entropy maximization Detailed composition Species identity Mole or mass fractions MAX S N M N M ( y i ) = y i ln( y i ) with i = 1 ± 20 Constraints properties from mixing More rules than (example): 100 unknown 1 y Mw mole fractions N d exp N = M i i = 1 di i i y Mw i Mw exp Mw = M i= 1 y i i i = 1 y i = 1 S.P. Pyl et al., AIChE Journal, 56, 12, ,

12 SIMCO results: Hydrocarbons n-paraffins i-paraffins Naphthenes Aromatics 12

13 Outline Introduction Feedstock: renewables Kinetics Reactor Process Conclusions 13

14 Thermochemicalconversionof biomass GAS BIO-OIL CHAR CO H 2 NH 3 CO CH 2 4 H 2 O HCN C 2H 4 5wt% 35wt% 13wt% 85wt% Torrefaction Slow pyrolysis Fast pyrolysis Gasfication 20wt% 30wt% 75wt% 5wt% 75wt% 35wt% 12wt% 10wt% LIGNOCELLULOSIC BIOMASS Toraman, H.E.et al.,bioresource Technology, 207, ,

15 Model componentsforbiomasspyrolysis H 2 O CH 4 H 2 C 2 H 4 CO CO 2 structural moieties found Study molecules with in biomass For example: T 15

16 Reaction Gamma-ValeroLactone(GVL) families pyrolysis 1) Scission 2) Hydrogen abstraction 3) β-scission/addition 4) Concerted ring opening De Bruycker, R. et al., Combustion and Flame, 164, , 2016 De Bruycker, R. et al.,proceedings of the Combustion Institute, 35, ,

17 Outline Introduction Feedstock Kinetics Reactor Process Conclusions 17

18 Goal CRACKSIM: steamcrackingkinetics Reactor geometry Operating conditions Feedstock properties Simulation model Product specs Modeling Microkinetic model:cracksim Reactor model Detailed feedstock composition Fundamental reactor model Detailed product composition 18

19 Outline Introduction Feedstock Kinetics: reaction network Reactor Process Conclusions 19

20 Families of elementary reactions Bond dissociation and radical recombination R 1 R 2 R 1 + R 2 Hydrogen abstraction (inter- and intramolecular) R 1 H + R 2 R 1 + R 2 H Radical addition and β-scission (inter- and intramolecular) R 1 + R 2 R 3 R 1 R 2 R 3 20

21 Decomposition scheme: n-hexane RH β species PSSA toµradicals 21

22 µandβnetworks: CRACKSIM CRACKSIM β network Bi- and monomolecular reactions for β radicals R 1 R 2 R 1 + R 2 R 2 + R 2 H R 1 H + R 1 R 1 + R 2 R 3 R 1 R 2 R 3 µ network Monomolecular reactions for µ radicals R 1 R 1 R 2 R 1 + R 2 R 2 -R 1 H + R 2 R 3 R 1 -R 2 R 3 H R 2 -R 1 R 1 R 2 R 3 R 2 R 1 R reversible reactions 51 molecules 43 radicals schemes 676 molecules 22

23 Validation Experimental yields during steam cracking of bio-derived hydrocarbons Feedstock composition: MW average = 230g/mol 51wt% normal alkanes 49wt% branched alkanes F HC,0 = 0.04 g/s, F H2O,0 = 0.02 g/s, P = 0.17 MPa calculated 23

24 Network generators RING Daoutidis Minneapolis MECHEM Valdes-Perez Pittsburgh Genesys PRIM(-O) ReNGeP LCT, Ghent, Belgium RAIN Ugi Munich, Germany REACTION Blurock Lund, Sweden RNG Lederer Prague, Czech Republic CASB Porollo Josjkar-Ola, Russia RDL Mavrovouniotis Evanston COMGEN Truong Salt Lake City RDL++ Vankatasubramanian West Lafayette NetGen Broadbelt Delaware EXGAS Battin-Leclerc Nancy, France RMG Green Cambridge MAMOX(+)(++) Ranzi Milan, Italy KING Di Maio Cosenza, Italy MECHGEN Héberger Budapest, Hungary GRACE, Yoneda KUCRS, Miyoshi Tokyo, Japan 24

25 A new program for kinetic model construction GENESYS - GENEration [of reacting ] SYStems Genesys: Kinetic model construction using chemo-informatics Vandewiele, N.M.; Van Geem, K.M.; Reyniers, M.-F.; Marin, G.B. Chemical Engineering Journal, , , 2012 Van de Vijver, R. et al. International Journal of Chemical Kinetics, 47 (4), ,

26 Graph theory yields powerful algorithms Pattern Recognition Reactant -> product Molecule Identifiers Kinetics of Chemical Reactions : Decoding Complexity G.B. Marin and G.S. Yablonsky,Wiley-VCH Verlag, 446 pages, 2011 Vandewiele, N.M. et al., Journal of Computational Chemistry, 36 (3), ,

27 Outline Introduction Feedstock Kinetics: ab initio Reactor Process Conclusions 27

28 Objective: data base Reactor simulation for hydrocarbon radical chemistry Reactor model Solver Reaction network Kinetic and thermodynamic data Develop a consistent data set based on ab initio calculations? Larger using species group additivity Van de Vijver, R. et al., Chemical Engineering Journal, 278, ,

29 D A X C B Notation: X-(A)(B)(C)(D) Benson s group additive method Benson group X: Central atom Valence 2 { C, C d, O, CO, CCO, C, C d } A, B, C, D: Ligand H, C, C d, O, CO, CCO, C, C d, O,CO,CCO Group additivity for thermochemistry f = + i GAV f ( groupi ) NNI j j f = f H, S o int, C o p S o int = S + σ R ln( n opt ) Group additive values (GAV) Corrections for non-nearest-neighbor interactions (NNI) hydrogen bonds gauche interactions other interactions 29

30 From small to large species with group additivity 2-methoxy-2-methylbut-3-enoic acid O O O OH O Additivity OH O Atoms in large molecules Group additivity Atoms in small molecules with similar surroundings additive groups C d -(H) 2 C d -(C)(H) C-(C)( C d )(O)(CO) O-(C) 2 C-(O)(H) 3 C-(C)(H) 3 CO-(C)(O) O OH = O-(CO)(H) 30

31 Outline Introduction Feedstock Kinetics: thermo first Reactor Processen Conclusions Ince, A.et al. AIChE Journal, 61 (11), ,

32 Thermo e.g. for oxygenates: GAV data base Database of thermodynamic data, f H, S and C p (300 K-1500 K) 450 oxygenate compounds CBS-QB3 methodology 1D-HR approximation for all internal rotors 2-hydroxy-2-methyl-propanal 157 GAVs using Benson s GA method 26 NNI corrections (mainly hydrogen bonds) 77 HBIs for the thermochemistry of radicals NNI8 Hbr_H-O-C-CO 32

33 Outline Introduction Feedstock Kinetics: rate coefficients Reactor Process Conclusions 33

34 Kinetics: computational approach Conventional Transition State Theory (high pressure limit) A + B [AB] C Transition state k kbt (T ) = κ(t ) h q q A q B V m e E RT 0 Products E electronic Electronic barrier E 0 The CBS-QB3 ab initio method is used. Partition functions q Ideal gas approximation Hindered Rotor (1D-HR) Reactants Reaction coordinate Tunneling coefficient κ Eckart E 34

35 Group additivityforkinetics:database of GAV o Transition state for hydrogen abstraction K 5 K 4 X 2 C 1 H C 2 Y 2 K 6 K 2 K 1 X 1 K 3 L 2 Y 1 L 3 L 4 K 7 X 3 K 8 L 7 Y 3 L 8 L 1 L 6 L 5 k Arrhenius equation k = Ae E a RT = κ n k = κn Eckart e GA e ~ Ae Ea RT n e Number of single events = j n opt, n opt, j σ j j σ K 9 L 9 E a log A ~ ( T ) ~ Group additivity for Arrhenius parameters Primary Secondary Tertiary = o ο E a,ref ( T ) + GAV E ( C ) + GAV a E ( ) + i X a i i = 1 i = 1 i = 1 ( T ) GAV ( Y ) + E = o ο log Aref ( T ) + GAV ( Ci ) GAV ( X i ) A ~ + log loga ~ + i= 1 i= 1 i= 1 GAV ο loga ~ ο E a ( Y i i o E ~ ) + log A o res a, res Proposed by Saeys et al. for activation energies (AIChE J. 2004, 50 (2), ) Extended by Sabbe et al. for pre-exponential factors (Phys. Chem. Chem. Phys. 2010, 12, ) 35

36 Kinetics: data base of GAV o Hydrogen abstraction (ethyl + ethenyl methylether) Paraskevas, P. et al.,journal of Physical Chemistry A, 119 (27), , 2015 O H 2 C CH 3 + C H 3 CH 2 CH 3 H C 1 H C 2 O H H H CH=CH 2 H 3 C CH 3 + H 2 C O CH 2 E o o o o a T ) = Ea,ref ( T ) + GAV E ( C1) + GAV E ( C 2) + GAV E ( X 1) + GAV E ( Y 1) ( a a a a + E a, res CH 3 + CH 4 C 1 -(C)(H) 2 C 2 -(O)(H) 2 C-(C 1 )(H) 3 O-(C 2 )(C d ) H kjmol kjmol kjmol kjmol kjmol kjmol -1 E a, ab initio= 62.0 kjmol -1 E a,ga = 62.6 kjmol -1 CH 3 + CH 4 C 1 -(C)(H) 2 C 2 -(O)(H) 2 C-(C 1 )(H) 3 O-(C 2 )(C d ) H log(a GA /m 3 mol -1 s -1 )= loga ab initio =

37 Kinetics: validation of group additivity K logk GA H additions H β scission C addition C β scission H-abstraction log(k AI /m³mol -1 s -1 ;s -1 ) Sabbe et al. ChemPhysChem 2008, 9 (1), Sabbe et al. ChemPhysChem, 2010, 11(1), Sabbe et al. PhysChemChemPhys, 2010, 12 (6),

38 Outline Introduction Feedstock Kinetics Reactor Process Conclusions 38

39 Outline Introduction Feedstock Kinetics Reactor : pilot scale Process Conclusions 39

40 SteamCrackingPilot Plant High temperature sampling system Gas-Fired Furnace + Reactor Online Analysis Section H 2 O HC Feed Control Room 40

41 Process conditions Pilot results: C 1 /C 2 /C 3 /C 4 Feed 3 wt% C 1 67 wt% C 2 22 wt%c 3 8 wt% C 4 COT K without a single adjusted parameter COP MPa Ab initio predicted yield (wt%) Steam dilution methane ethene ethane propene propane 0 kg/kg Ab initio predicted yield (wt%) ,3-butadiene 1-butene n-butane benzene H Experimental yield (wt%) Experimental yield (wt%) 41

42 Outline Introduction Feedstock Kinetics Reactor: 3D alternatives Process Conclusions 42

Deposition of a carbon layer on the reactor surface Thermal efficiency Product selectivity Decoking procedures Coke formation Estimated annual cost to industry: $ 2 billion Mitigation by - Feed

43 Deposition of a carbon layer on the reactor surface Thermal efficiency Product selectivity Decoking procedures Coke formation Estimated annual cost to industry: $ 2 billion Mitigation by - Feed additives - Metallurgy & surface technology - 3D reactor technology 2 L. Benum, "Achieving Longer Furnace Runs at NOVA Chemicals," in AIChESpring National Meeting, 14th Annual Ethylene Producers Conference, New Orleans, Louisiana,

44 Coke formation: 3D reactor technologies Cokes formed here T coking rate Reduce convective heat resistance Increase surface area Better mixing ~ 15 C ~60 ~100 C C *Borealis.com, kubota.com, Technip.com 44

Thereain tnosuchthingas a free lunch 3D reactor technology Flow Chemistry Increased heat transfer 3D CRACKSIM Computer Models Increased friction Reduced coking

45 Thereain tnosuchthingas a free lunch 3D reactor technology Flow Chemistry Increased heat transfer 3D CRACKSIM Computer Models Increased friction Reduced coking rate Quantify coking & selectivity effect Higher average reactor pressure Longer run length More capacity Economics Selectivity * Milton Friedman - TANSTAAFL 45

66th Canadian Chemical Engineering Conference Sustainability & Prosperity October 16-19 (2016) Québec City (Canai

46 66th Canadian Chemical Engineering Conference Sustainability & Prosperity October (2016) Québec City (Canai Helicoidalfinnedtubes D e α Pitch Radiant coils in Borealis Furnace, KBR Radius Computational domain: 1 fin with periodic boundaries 46

47 Outline Introduction Feedstock Kinetics Reactor : CFD models Process Conclusions 47

(LES) Resolve relevant energy containing scales, model the smaller energy

48 Resolving turbulence Reynolds-Averaged Navier-Stokes (RANS) Single model for all scales, additional equations to provide closures Large Eddy Simulation (LES) Resolve relevant energy containing scales, model the smaller energy dissipating eddies Direct Numerical Solution (DNS) Fully resolve all time and length scales 48

49 Radialtemperatureprofiles Bare Straight Helix SmallFins Optimized More uniform gas temperature Lower metal temperature 49

50 Axialwalltemperatureandcokingprofile - 50 C Cracking reaction model 26 components 13 radical species 212 elementary reactions - 49% Coke model Coking model S. Wauters and G.B. Marin, IEC Research, 41, , 2002 S. Wauters and G.B. Marin; Chemical Engineering Journal, 82, ,

Effect on start of run yields - 0.7 wt% + 0.

51 Effect on start of run yields wt% wt% Radical reaction model 26 components 13 radical species 212 elementary reactions 51

52 Outline Introduction Feedstock Kinetics Reactor: run length Process Conclusions 52

53 Run lengthsimulation Increasing run length SOR 48h 96h

54 Millisecond propane cracker Feedstock kg/h propane Propane conversion % (± 0.05%) Steam dilution kg/kg CIT C COP 170 kpa Different geometries simulated Same reactor volume Same axial length Same minimal wall thickness Bare Fin c-rib 54

55 Non-uniform coke layergrowth Fin c-rib SOR (0 hrs) SOR 48 hrs 10 days 55

56 Tube Metal Temperature Bare Fin c-rib Thermal resistance coke layer Max. TMT increases 56

57 Pressuredrop Bare Fin c-rib Cross-sectional flow area decreases due to coke Pressure drop increases 57

58 Outline Introduction Feedstock Kinetics Reactor Process Conclusions 58

59 Outline Introduction Feedstock Kinetics Reactor Process: fire box Conclusions 59

60 Coupledreactor-furnacesimulation External coil temperature Reactor (COILSIM1D) convergence Furnace simulation FURNACE/ COILSIM1D Heat flux to reactors Hu, G. et al.,industrial & Engineering Chemistry Research, 54 (9), ,

61 Full furnace simulation Ultra Selective Conversion (USC) 100% floor burner Fuel composition mol%: CH 4 (89%)- H 2 (11%) U coil Feedstock: Naphtha Coupled modeling 3D CFD furnace model 1D reactor model (COILSIM1D) Detailed cracking kinetics (CRACKSIM) Zhang, Y. et al.,aiche Journal 61 (3), ,

62 Detailed : long flame burners detailed simplified 62

$Methane mole fraction detailed$

63 Fluegas: velocityandconcentrationfields Methane mole fraction detailed simplified Detailed case Stronger turbulence Faster reaction 63

64 Tube wall temperaturefield: local hot spots This information is key for control 64

65 Outline Introduction Feedstock Kinetics Reactor Process: convection section Conclusions 65

overheater-2 To radiation section Flue gas in ~ 1400 K Schematic of convection section

66 Steamcracker convectionsection Flue gas out ~ 700K Heavy feed Mixing nozzle Feed Evaporator Steam super heater Feed Evaporator Mixture overheater-1 Steam Mixture overheater-2 To radiation section Flue gas in ~ 1400 K Schematic of convection section Nozzle Feed-steam mixture overheater-1 Mahulkar, A.V. et al., Chemical Engineering Science, 110, 31-43,

Gas condensate: multicomponent mixture Stick 1200 Splash Rebound Normal Weber number 1000 800 600 400 200 Stick Splash Limited Splash Rebound 0 480 530 580 630 680 730 780 T LF Wall temperature (K)

67 Gas condensate: multicomponent mixture Stick 1200 Splash Rebound Normal Weber number Stick Splash Limited Splash Rebound T LF Wall temperature (K) 820 crit We Norm Limited Splash Wider stick regime as compared to that in single component droplet regime map For splash and limited splash, the no. of daughter droplets formed is greater than predicted by correlations available in literature Mahulkar, A.V. et al., Chemical Engineering Science, 130, ,

68 Outline Introduction Feedstock Kinetics Reactor Process: hot section Conclusions 68

69 The COILSIM1D package Transfer Line Exchanger Convection section Reactor Coil Wall Burners Firebox Long Flame Burners 69

70 The COILSIM1D package Upstream simulators Downstream simulators... 70

71 Outline Introduction Feedstock Kinetics Reactor Process Conclusions 71

72 Conclusions Shannon Entropy maximization to reconstruct feedstocks in terms required for a microkinetic model Consistent data set for thermochemistry and kinetics of hydrocarbon and oxygenates radical chemistry Ab initio simulation of steam cracking of C 2 /C 3 /C 4 and oxygenates Compatibility of renewable feedstocks with existing plants Emergence of 3D reactor technologies based on CFD Integration of reactor/convection section/furnace integration of computational chemistry methods with engineering tools at larger time and length scales and experimental validation provides a powerful tool for the optimization and/or design of industrial units 72

73 Acknowledgments The Long Term Structural Methusalem Funding 73

74 Acknowledgments Profs. Marie-Françoise Reyniers, Geraldine Heyndericks Maarten Sabbe, Tony Arts, Feng Qian Drs. Steven Pyl, Amit Mahulkar, Nick Vandewiele, Ruben Debruycker, Carl Schietekat, Thomas Dijkmans, Paschalis Paraskevas, Marco Djokic, Andres Munoz PhD students David Van Cauwenberghe, Pieter Reyniers, Brigitte Devocht, Ruben Van de Vijver, Laurien Vandewalle, Nenad Ristic, Alper Ince, Ezgi Toraman, Yu Zhang, Marco Virgilio 74

75 People

76 Glossary SFT: Swirl Flow Tube, a tube with a helicoidal centerline. The helix amplitude is smaller than or equal to the tube radius. Swirl flow: a whirling or eddying flow of fluid. Swirl number: ratio of tangential over axial momentum transfer Wall shear stress: component of stress parallel with the wall. It is the product of the viscosity and the derivative of axial velocity with respect to the radial coordinate. 76

Oxygen-Containing Contaminants and Steam Cracking: Understanding their Impact Using COILSIM1D

Oxygen-Containing Contaminants and Steam Cracking: Understanding their Impact Using COILSIM1D Ismaël Amghizar 1, Andrés E. Muñoz G. 2, Marko R. Djokic 1, Kevin M. Van Geem 1, Guy B. Marin 1 1 Laboratory