Dynamic Modeling and Optimization of Large-Scale Cryogenic Processes Maria Soledad Diaz Planta Piloto de Ingeniería Química Universidad Nacional del Sur -CONICET Bahía Blanca, ARGENTINA Outline Objective Natural Gas Processing Plants Methodology Mathematical Models Discussion of Results Conclusions and Current Work 2 Objective Dynamic optimization model for natural gas processing plant units • • • • • • Dynamic energy and mass balances Phase equilibrium calculations with cubic equation of state Hydraulic correlations Phase change in countercurrent heat exchanger Carbon dioxide precipitation in column Boiler dynamic optimization for controllers parameters Simultaneous dynamic optimization approach • Discretization of state and control variables • Resolution of large-scale Nonlinear Programming problem Analysis of main variables temporal and spatial profiles 3 Natural Gas Processing Plants Provide ethane as raw material for olefin plants (ethylene) and petrochemical Technology: Turboexpansion High pressure Cryogenic conditions 4 NATURAL GAS PLANT D E H Y D R A T I O N Ethane To Ethylene Plant Ethane Treatment CO2 TRAIN A TRAIN B C2 C3 C4 Propane Compression/ Recompression D E H Y D R A T I O N TRAIN C Butane Separation Fractionation Gasoline To Recompression TRAIN A C3 C4 C3 C4 TRAIN B Pipelines Absorption Plant 5 Natural gas composition Component Feed A Feed B Nitrogen 1.44 1.37 Carbon dioxide 0.65 1.30 Methane 90.43 89.40 Ethane 4.61 4.43 Propane 1.76 2.04 Butanes 0.77 0.96 Pentanes+ 0.34 0.50 6 Cryogenic Sector Turboexpander Demethanizer Cryogenic Heat Exchangers High Pressure Separator to deethanizer 7 column Previous Work Operating conditions optimization in natural gas processing plants (NLP) (Diaz, Serrani, Be Deistegui, Brignole, 1995) Automatic design and debottlenecking of ethane extraction plants (MINLP) (Diaz, Serrani, Bandoni Brignole, 1997) Thermodynamic model effect on the design and optimization of natural gas plants (NLP) (Diaz, Zabaloy, Brignole, 1999) Flexibility of natural gas processing plants - Dual mode operation (Bilevel NLP) (Diaz, Bandoni, Brignole, 2002) 8 Dynamic Optimization Problem General Formulation min (z(tf),y(tf), u(tf), tf , p) z(t),y(t), u(t), tf , p st F(dz(t)/dt ,(z(t), y(t), u(t), p, t)=0 G(z(t) , y(t), u(t), p, t) = 0 zO = z(0) z(t) [zl, zu] u(t) [ul, uu] t, time z, differential variables y, algebraic variables , y(t) [xl, xu] , p [pl, pu] tf, final time u, control variables p, time independent parameters 9 Dynamic Optimization Strategy Simultaneous Approach Cervantes, Biegler (1998) Nonlinear DAE optimization problem Discretization of Control and State variables Collocation on finite elements Large-Scale Nonlinear Programming Problem (NLP) Interior Point Algorithm Biegler, Cervantes, Waechter (2002) 10 Simultaneous Dynamic Optimization Discretization of differential equations (DAEs) and state and control variables Large-scale nonlinear problem Profile constraints handled directly Direct incorporation of equipment design variables DAE model solved only once Converges for unstable systems 11 Nonlinear Programming Problem Application of Barrier method n min f ( x) ln s j j 1 s.t c( x) 0 s.t c( x) 0 x0 sx0 As 0, x*() x* 12 Cryogenic Heat Exchangers and HP Separator To Recompression Rodriguez, Bandoni, Diaz (2004) Residual Gas V To Turboexpander ht Phase Change To Demethanizing Column Horizontal High Pressure Separator Ts2_i = 322 K Natural Gas 13 HE Model (no Phase Change) Energy Balances: Partial Differential Equations System Tube side ht * Asup Tt( z ,t ) Tt( z ,t ) Ts( z ,t ) Tt( z ,t ) vt t z t * Cpt * At * L Shell side hs * Asup Ts( z ,t ) Ts( z ,t ) Tt( z ,t ) Ts( z ,t ) vs t z s * Cps * As * L Spatial Discretization: Finite Differences ODE System 14 HE Model (no Phase Change) Energy Balance at cell i ht * Asup dTti vti Tsi Tti Tti1 Tti dt z ti * Cpt * At * L Tube side hs * Asup dTsi vsi Tti Tsi Tsi1 Tsi dt z si * Cps * As * L Shell side Hot stream at Ts0 Ts1 Ts2 Ts3 Ts4 Ts5 Ts6 Tt1 Tt2 Tt3 Tt4 Tt5 Tt6 Ts1 Ts2 Ts3 Ts4 Ts5 Ts6 Multicell model E-1 15 Cold stream at Tt0 Algebraic Equations (no Phase Change) z j ,i A j B j * T j ,i j ,i To Recompression Residual Gas M*P z j ,i * R * T j ,i v j ,i Fj j ,i * A j G i = 1, …, N (cells) j = tube or shell side HE model DAE System Natural Gas 16 HE Model (Partial Phase Change on Shell Side) Rodriguez, Bandoni, Diaz (2005) Energy Balance at cell i dEi Lpi*hpi + Vpi*Hpi - Li*hi - Vi*Hi + Qit dt Algebraic Equations at cell i Ki , j iL, j V i , j hi h ideal i Hi H yi , j K i , j xi , j hi ideal i H i Qt ,i U 0 * A* Tml ,i hiideal H iideal yi , j xi , j 0 j j nc hiideal ( Ti )xi , j ,j j 1 nc H iideal ( Ti )yi , j ,j j 1 17 HE Model (Partial Phase Change on Shell Side) Compressibility factor ziV zV ( Tsi , Psi , yi ) ziL z L ( Tsi , Psi , xi ) Residual enthalpies H i H ( Tsi , Psi , yi ) Soave-Redlich-Kwong EoS hi h( Tsi , Psi , xi ) Fugacity coefficients iV, j V ( Tsi , Psi , yi ) iL, j L ( Tsi , Psi , xi ) Psi siL * ziL * R * Tsi PmolsiL Psi siV * ziV * R * Tsi PmolsiV 18 HE Model (Partial Phase Change on Shell Side) Bell-Delaware method Pressure drop Pi 2 V ( i )* Vmax (i ) Pc ( i ) Ka N c * Kf ( i )* 2 Ps (i) A * Pc (i) B * Pw (i) * Rl C * Pc (i) Pw (i) 2 0.6 * N cw * m T2 (i) Kf (i) A V (i) B C D E Re(i) Re(i) 2 Re(i)3 Re(i) 4 V (i) * Vmax (i) * D0 Re(i) 19 High Pressure Separator dM Lp+ Vp - L - V dt Horizontal tank 2 M D VolV in Long L L 2 L Cv Ps Pt ht r M V V VolV ML M MV M L L Long 2 ht 2 2 2 r r arccos h r h t t PmolL r 20 Turboexpander Feed: vapor from High Pressure Separator Polytropic expansion (η = 1.26, natural gas) 1 Ts Ps Te Pe Assumption: Static mass and energy balances Thermodynamic predictions: Soave-Redlich-Kwong Algebraic equations: Same as Flash Outlet partially condensed stream: to Demethanizing column 21 Demethanizing Column Diaz, Tonelli, Bandoni, Biegler (2003) Differential Equations at stage i (1 ≤ i ≤ N) dM i Fi Vi1 Li1 Vi Li dt dmij dt Fi zi , j Vi1 yi1, j Li1 xi1, j Vi yi , j Li xi , j j=1,…,ncomp dEi Vi1 H i1 Li1hi1 Vi H i Li hi Fi i H Fi 1 i hFi Qsri dt N = 8; ncomp = 10 22 Demethanizing Column Algebraic Equations Compressibility factor ( z iV , z iL) Residual enthalpies (Hi, hi) Fugacity coefficients (i , j , i , j) V Ki , j iL, j V i , j SRK L yi , j K i , j xi , j yi , j xi , j 0 j j H i H iideal H i hi hiideal hi 23 Demethanizing Column Algebraic Equations 2 Vapor volume M iL Di V Voli H plate ,i L i 2 Vapor and liquid holdup M iV iV VoliV M i M iV M iL Component holdup mi , j M iV yi , j M iL xi , j Internal energy holdup Ei M iV ( Hi pi V i ) M iL ( hi pi L i ) 24 Demethanizing Column Liquid holdup and hydraulic correlations 2 D M iL 0.896 i Hwi Hwoi iL 2 Vi 4 Hoi 5.56.10 V i AholeiCoi 2 iV PmoliV L L i Pmoli Li L i Weirli 2/ 3 Hwoi 300.01495 Fwi 2/ 3 Hldropi Hwi Hwoi Pressure drop P1 PTOP KK1V V12 Pi Pi 1 0.24917 .10 5 Hoi Hldrop i iL PmoliL 25 Demethanizing Column Carbon dioxide solubility constraints Phase Equilibrium iV,CO 2 iL,CO 2 iS,CO 2 fiV,CO 2 fi ,LCO 2 fi S,CO 2 Assumption No hydrocarbon in solid phase To avoid CO2 precipitation fi S,CO 2 fi S,CO 2 f iV,CO 2 f i ,SCO 2 fi ,LCO 2 fi S,CO 2 But fi ,LCO 2 fiV,CO 2 from VLE calculations at each stage 26 Demethanizing Column Carbon dioxide solubility constraints f iV,CO 2 f i ,SCO 2 Solid phase f i ,SCO 2 Pi ,SCO 2iV,CO 2 t Pi S,CO 2 TCO T T 2 14 .480 ln t i 65 .356 t i 1 ln t 14 .568 1 Ti TCO 2 TCO 2 PCO 2 T 2 T 3 47 .146 t i 1 14 .540 t i 1 TCO 2 TCO 2 Vapor phase f iV,CO 2 yi ,CO 2 Pi iV,CO 2 27 Demethanizing Column: Phase Existence Raghunathan, Diaz, Biegler (2004) Gibbs Free Energy minimization at stage i Karush Kuhn Tucker conditions Raghunathan, Biegler (2003) MPECs as additional constraints IPOPT-C under AMPL 28 Demethanizing Column: Phase Existence Relaxed Equilibrium Conditions yi , j i K i , j ( Ti , Pi , xi , j , yi , j )xi , j i 1 siV siL siL M iL 0 siV M iV 0 1 M iL 0 , M iV 0 1 M iL 0 , M iV 0 1 M iL 0 , M iV 0 , siV 0 , siL 0 29 Demethanizing Column: Phase Existence Liquid Holdup and Hydraulic correlations 2 Di L M i 0.896 Hwi iL M iL M iL 2 2 D M iL 0.896 i Hwoi iL 2 Hwoi 300.01495 2/ 3 L Fwi L i i Weirli 2/ 3 3 L 2 Li kd ( M i ) M iL M iL 0 M iL 0, M iL 0 30 Optimization Problem: HEs + HPS min 0tf T TSP dt 2 T st . h t L G DAE System 60 L 80( kmol / min) 5 G 80( kmol / min) Tout 303 Tout 306( K ) Alg. Eqns = 403 Differ. Eqns = 37 (10+8 cells in HEs) 31 Numerical Results: HEs + HPS Optimization variables: G, L 20 elements 65.330 1.00 2 collocation points 65.325 Lt 0.95 h 0.90 0.85 18274 disc. variables 0.80 0.75 65.315 0.70 0.65 65.310 0.60 0.55 65.305 0.50 0.45 65.300 0 5 10 15 20 25 0.40 30 time (min) 32 h (m) 18 iter. (3 barrier problems) Lt (Kmol/min) 65.320 Numerical Results: HEs + HPS Optimization variables: G, L 83 307 82 G0 81 Tout 306 80 305 78 304 77 76 303 Tt (K) Temperature vs time 212.0 75 74 302 211.9 73 72 301 211.8 71 70 0 5 10 15 time (min) 20 25 300 30 Ts (K) G0 (Kmol/min) 79 211.7 211.6 211.5 211.4 0 5 10 15 20 25 time (min) 33 30 Numerical Results: HEs + HPS Temp. profile, tubes side Temp. profile, shell side (no phase change) 34 Numerical Results: HEs + HPS 58.5 35.0 58.2 30.0 57.9 57.6 25.0 57.3 Ps (bar) 20.0 57.0 L (Kmol/min) 56.7 15.0 56.4 10.0 56.1 55.8 5.0 55.5 0 2 time (min) 11.5 21 30.5 40 1 2 3 4 5 6 cells Pressure profile, shell side 6 cells 0.0 5 4 3 2 1 0 2 11.5 21 30.5 40 time (min) Condensed fraction, shell side 35 Optimization Problem: Demethanizing Column min tf 0 ethane SP dt 2 st . DAE System 15 PTOP 22( bar ) 99 QREB 200( kJ / min) 0 xB ,CH 4 0.008 f iV,CO 2 0.90 f i ,SCO 2 Alg. Eqns = 385 Differ. Eqns = 97 36 Numerical Results: Demethanizer Feed A: Low CO2 content (0.65%) 20 82 Optimization variable: PTOP 80 2 collocation points P Ethane recovery 18 Ptop (bar) 20 finite elements 76 74 21357 discretized variables 40 iter. (3 barrier problems) 16 72 ss = 80.9 % 70 0 10 20 30 40 50 Time (min) 37 Ethane recovery (%) 78 Numerical Results: Demethanizer Feed B: High CO2 content (2%) Optimization variable: PTOP Ptop (bar) 2 collocation points 84 80 P Ethane Recovery 18 43 iter. (3 barrier problems) 76 ss = 74.5 % 16 72 0 10 20 30 40 Time (min) 38 Ethane recovery (%) 20 finite elements 20 Numerical Results: Demethanizer Feed B (high CO2 content) 84 No solubility constraints P Ethane recovery 20 20 finite elements Ptop (bar) 80 78 18 76 2 collocation points 59 iter. (4 barrier problems) ss = 76.6 % 74 16 0 10 20 30 40 50 Time (min) 39 Ethane recovery (%) Optimization variable: PTOP 82 Numerical Results: Demethanizer Feed A Optimization variable: 178 Reboiler Heat Duty (QR) 57 iter. (5 barrier problems) ss = 77.8 % (PTOP=19bar) Reboiler Heat Duty (MJ/min) 2 collocation points 176 Active constraint: Methane mole fraction in bottoms 0.007 174 QREBOILER XCH4 172 0.006 170 0.005 168 166 0.004 164 162 0.003 0 5 10 15 20 25 30 35 40 Time (min) 40 Bottom methane mole fraction 20 finite elements 0.008 Numerical Results: Demethanizer Start Up MPECs Optimization variable: Three time periods Ideal Gas, Ideal solution Ethane Recovery PTOP, QR Solved in AMPL with IPOPT-C (Raghunathan, Biegler, 2003) Time (min) 41 Numerical Results: Demethanizer Start Up MPECs PTOP, QR 3rd time period 17944 discretized variables 576 complementarity constraints Reboiler Heat Duty (MJ/min) Optimization variable: 97 iterations Time (min) 42 Boiler Dynamic Optimization Rodriguez, Bandoni, Diaz (2005b) Dynamic optimization model to determine controller parameters Boiler Flue Gas High Pressure Steam Combustion chamber Feed Water Risers Superheater E-1 Air Steam drum Combustion Chamber Superheater RISER 1 RISER 2 Gas path Natural Gas Furnace Gas PI Controller (Liquid level in drum) Manipulated: Feed water flowrate 43 Boiler Dynamic Optimization DAE Optimization model Differential equations • Momentum and energy balances in risers and superheater • Mass and energy balances in drum • Integral part of controller and valve equation Algebraic equations • Friction loss • Mixture properties • Vapor holdup in drum • Gas – temperature distribution equations in furnace, risers ans superheaters (6 eqns.) • Air / fuel ratio • Drum geometric relations (5 eqns.) • Steam tables correlations for saturated and high pressure steam density, pressure and enthalpy 44 Boiler Dynamic Optimization DAE Optimization model Optimization variables • PI controller parameters Step change in high pressure steam demand 6 32 4 Liquid Level Drum (ft) Feed Water Flowrate (lb/s) 30 28 26 24 2 0 -2 22 -4 0 200 400 600 800 1000 0 200 400 Time (s) 600 800 1000 Time (s) Additional PI controllers (current work) • Outlet steam pressure (fuel gas flowrate) • Superheated steam temperature (air excess) 45 Conclusions Rigorous dynamic optimization models for cryogenic train in Natural Gas Processing plant within a simultaneous approach Thermodynamic models with cubic equation of state Carbon dioxide solubility addressed in both liquid and vapor phases as path constraints, at each column stage Phase detection through the addition of MPECs for Gibbs free energy minimization Boiler dynamic model for controller parameters Simultaneous approach provides efficient framework for the formulation and resolution of DAE optimization problems for large-scale real plants 46 References Biegler, L. T., A. Cervantes, A. Waechter, “Advances in Simultaneous Strategies for Dynamic Process Optimization,” Chem. Eng. Sci., 57, 575-593, 2002 Cervantes, A., L. T. Biegler, “Large-Scale DAE Optimization using Simultaneous Nonlinear Programming Formulations”, AIChE Journal, 44, 1038, 1998 Diaz, S., A. Serrani, R. de Beistegui, E. Brignole, "A MINLP Strategy for the Debottlenecking problem in an Ethane Extraction Plant"; Computers and Chemical Engineering, 19s, 175-178, 1995 Diaz, S., A.Serrani, A. Bandoni, E. A. Brignole, “Automatic Design and Optimization of Natural Gas Plants", Industrial and Engineering Chemistry Research, 36, 2715-2724, 1997 S. Diaz, M. Zabaloy, E. A. Brignole, “Thermodynamic Model Effect on the Design And Optimization of Natural Gas Plants”, Proceedings 78th Gas Processors Association Annual Convention, 38-45, March 1999, Nashville, Tennessee, USA Diaz, S., A. Bandoni, E.A. Brignole “Flexibility study on a dual mode natural gas plant in operation”, Chemical Engineering Communications, 189, 5, 623641, 2002 47 References Diaz, S., S. Tonelli, A. Bandoni, L.T. Biegler, “Dynamic optimization for switching between steady states in cryogenic plants”, Foundations of Computer Aided Process Operations, 4, 601-604, 2003 Diaz, S., A. Raghunathan , L. Biegler, “Dynamic Optimization of a Cryogenic Distillation Column Using Complementarity Constraints”, 449d, AIChE Annual Meeting, San Francisco, Nov. 16-19, 2003. Advances in Optimization Raghunathan, A., M.S. Diaz, L.T. Biegler, “An MPEC Formulation for Dynamic Optimization of Distillation Operations”, Computers and Chemical Engineering, 28, 2037-2052, 2004 M. Rodríguez, J. A. Bandoni, M. S. Diaz, “Optimal dynamic responses of a heat exchanger/separator tank system with phase change”, submitted to Energy Conversion and Management, 2004 Rodríguez, M., J. A. Bandoni, M. S. Diaz, “Dynamic Modelling and Optimisation of Large-Scale Cryogenic Separation Processes”, IChEAP-7, The Seventh Italian Conference on Chemical and Process Engineering, 15 - 18 May 2005a Giardini Naxos, Italy Rodríguez, M., J. A. Bandoni, M. S. Diaz, “Boiler Controller Design Using Dynamic Optimisation”, PRES05, 8th Conference on Process Integration, Modelling and Optimisation for Energy Saving and Pollution Reduction, 15 48 18 May 2005b Giardini Naxos, Italy Appendix: Soave Redlich Kwong Equation of State z Compressibility factor L V for a mixture ( zi , zi ) V a V b RgT ( V b ) a x*j z*j ; x*j x j a j j ; z j x*k ( 1 k j ,k ) k b b j ; b j 0.08664 RgTcj / Pcj j 0.42747 aj Pcj Rg2Tcj2 0.5 T 0.5 ; j 1 m j 1 Tcj Soave modification (Temperature dependence) m j 0.48 1.574 j 0.176 2j 49 Appendix: Soave Redlich Kwong Equation of State Fugacity Coeff. for component j in mixture ( iV, j ,iL, j ) * 2 a z bj B A j j j ln j ( z 1 ) ln( z B ) ln1 b B a b z bj H 1 Residual enthalpy B * * 1 m j 1 ln1 x j z j in mixture ( H i , hi ) RgT bRgT zj j A aP bP ; B Rg2T 2 RgT 50