Eduard Manek,Juma Haydary*

Institute of Chemical and Environmental Engineering,Faculty of Chemical and Food Technology,Slovak University of Technology,Radlinského 9,812 37 Bratislava,Slovakia

Keywords:Residual hydrocracking Ebullated bed reactor Liquid recycle Aspen Plus Simulation

A B S T R A C T Oneofthecommercialmeanstoconvertheavyoilresidueishydrocrackinginanebullatedbed.Theebullatedbed reactor includes a complex gas-liquid-solid backmixed system which attracts the attention of many scientists and research groups.This work is aimed at the calculation of the internal recycle flow rate and understanding itseffectonother parametersoftheebullatedbed.Measureddatawerecollectedfroman industrialscaleresidual hydrocracking unit consisting of a cascade of three ebullated bed reactors.A simplified block model of the ebullated bed reactors was created in Aspen Plus and fed with measured data.For reaction yield calculation,a lumped kinetic model was used.The model was verified by comparing experimental and calculated distillation curves as well as the calculated and measured reactor inlet temperature.Influence of the feed rate on the recycle ratio(recycletofeedflowrate)wasestimated.Arelationbetweentherecycleflowrate,pumppressuredifference and catalyst inventory has been identified.The recycle ratio also affects the temperature gradient along the reactor cascade.Influence of the recycle ratioon the temperaturegradientdecreased with thecascade memberorder.

1.Introduction

Hydrotreating of residual oils in order to produce lower sulfur residual fuels came into prominence around 1960-1970.At the time,attention was focused on the production of low sulfur residual fuel oil.As a side effect,some low level conversion of the high boiling point(537°C+)material was observed.In general,the feedstock contained an appreciable amount of material with boiling point below 537°C,atmosphericresidueorlongresidue,whichdiffersfromvacuumresidue(short residue)typically containing only a low amount of distillates.In the early years,most of these hydrotreating units were fixed catalyst bed units and a predominant number of them were located in Japan.However,with the increased yield of the bottom of the barrel product from crude,more emphasis was placed on selective hydrogenation of the residue portion with an emphasis on the conversion of the residue material to distillates.Handling the heavy and viscous vacuum residue feedstock at higher residue conversion is limited by solid deposits and catalyst inventory activity.US patent 2987465 gives details on the process of contacting gas or liquid in a bed of random moving solids described as‘ebullated'.To maintain the ebullated state,liquid recycle of 19-27 volumes per volume of the charge stock is needed[1].

A schematic of an ebullated bed reactor is shown in Fig.1.Feed and gas enter at the bottom of the reactor and join a recycle oil stream.The mixture passes up through a distributor plate at sufficient velocity to expand the catalyst above the grid into a state of random turbulentmotion[2].The expandedbedis maintained slightly above thesettled level by closely controlling the recycle pump speed and monitoring the density by nuclear density meters[3].Recycle pump suction is supplied from the top of the reactor.A recycle pan is installed to disengage the gas before recycling the liquid[2].Top and bottom lines are provided for the routine addition and withdrawal of catalyst while the reactor is in operation.Temperatures are measured at the feed inlet,reactor plenum,reactor grid,and mixture outlet.Reactor temperature control is described in detail in[4].Additional temperature indicators are around the reactor wall and the internal thermo wells to monitor the ebullated bed and potential local hotspots.

The ebullated system has multiple variables which are difficult to measure directly.Mass or volume flow of the recycle stream is usually not measured in industrial scale units.Total amount of the catalyst inventory after several additional/withdrawal sequences,gas and liquid holdup in the reactor is not exactly known.Various approaches have been made to develop a kinetic model successfully describing the performance of heavy oil residue hydrocracking[5-10].Most of current residue models are focused on the description of the relation between input process parameters on the yield.Analyzing and optimizing the complexsetupofanindustrialreactorcascadeanddownstream separation section without laboratory experiments requires an efficientmathematical model of the unit.Simulation of the downstream separation section consisting only of a vapor-liquid separator,heat exchangers and fractionating columns can be developed in standard process simulators like PRO/II,Aspen Plus or Aspen HYSYS.Composition of the feed to the separation section can be obtained by combining distillation curves of all final products.This approach is limited only to the downstreamsectionanddoesnotprovidecontroloroptimizationofthereactor section.Proposing any technological changes leading to process optimization requires the knowledge of its effect on the internal streams parameters.However,these parameters are not measured in industrial scale units.Therefore a verified simulation model enabling thecalculationofflowandinternalstreamspropertiescanbeveryhelpful for the optimization of a hydrocracking unit.

Fig.1.Ebullated bed reactor for hydrocracking.

The aim of this work was to develop a model of industrial scale ebullated bed reactor cascade capable to calculate intrinsic parameters such as flow and properties of liquid recycle,inter-stage streams properties,and heat balance.This model can be used to investigate potential upgrades such as:implementation of inter-stage separators,quench materialoptimization,evaluationofthedependencebetweentheliquid recycling flow rate and the amount of catalyst in the reactor,or evaluation of the influence of increased throughput or conversion on the separation section.

2.Experimental

Fig.2.Simplified scheme of the hydrockracing unit.

Measured data for model development were obtained from database system of industrial residual hydrocracking plant for a three month continuous operation at various feed rates and reactor temperatures.Simplified scheme of the reactor section is shown in Fig.2.The feed is mixed with heavy aromatic oils,light cycle oil and the main column bottom from an FCC unit to suppress precipitation and subsequent coagulation of asphaltenes,sediments formation and to maintain the required hydrodynamic characteristics of the feed.Diluents represent up to 12 wt%of the total liquid feed mixture.Before entering the reactor cascade,preheated feed is mixed with high temperature hydrogen.The industrial reactor cascade consisted of three identical ebullatedbedreactorsfilledwithcommercialNi-Mo/Al2O3hydrocracking catalyst pellets.Temperature in the reactors was maintained at around 401-419°C and the pressure at 18-20 MPa.Effluents from thefirst to the second stage and from the second to the third stage were quenched by light hydrocarbon oil and hydrogen to maintain the temperature of the ebullated beds constant and also to compensate for the spent hydrogen.Effluent from the third reactor stage is subsequently depressurized and cooled down in the separator section.Hydrogen rich gas is then purified and recompressed as recycle gas.Liquid fractions are distilled into required products.In this paper,two basic cases(Table 1)were chosen for detailed performance comparison.Feed composition and diluent concentration were the same for both cases.Simulated distillation ASTM D2887((Table 2)was used to characterize the vacuum residue fraction.Simulated distillation ended at 98 vol%distilled and thus,the remaining 2 vol%of the heaviest portion was extrapolated.

Table 1 Operation parameters for the two exemplar cases

Table 2 Distillation curve of liquid feed

3.Simulation

Ebullated bed reactors are well mixed by a recycle pump and a flow distribution system.For the purpose of this work,the reactor series was considered as a cascade of continuous stirred-tank reactors.A kinetic model developed in a previous paper[8]was implemented in this work together with the Redlich-Kwong-Soave equation of state for phase equilibrium calculation.

The reaction pathways(Fig.3)of the kinetic model consist of primary vacuum residue(VR)cracking into offgasses(G),naphtha(GLN),kerosene(Ke),gas oil(GO)and vacuum gas oil(VGO),as well as of secondary cracking of vacuum distillates into gasoline,kerosene and gas oil.The proposed mathematical model consists of six ordinary differential equations based on the Guldberg-Waage law of mass action:primary cracking of vacuum residue(Eq.1),secondary cracking andproductionofvacuumdistillatesfromprimarycracking(Eq.2),production of gas oil,kerosene and naphtha fractions from primary and secondary cracking(Eq.3-5)and production of offgasses from primary cracking(Eq.6).Each lump is represented in the equations by its mass fraction.Kinetic constant as a function of temperature was described by the Arrhenius equation(Eq.7).Independent variables are:residence time(t),and reactiontemperature(T).Firstorder kineticswasassumed for all pathways.Feed rates were approximately calculated into residence times based on the volume flow and volume summation of the catalyst beds of all three reactors resulting in 64 min per reactor stage for Case 1 and 60 min per reactor stage for Case 2.

Fig.3.Pathways of vacuum residue hydrocracking.

where i represents reaction pathways 1-8.

The cascade of reactors was simulated in Aspen Plus environment.Measureddataformassflowsandtemperatureswereobtainedfromindustrial plant database,and composition of each stream was calculated by an Aspen Plus user defined model incorporating the kinetic model presented in our previous paper[8].

Single EBR reactor was represented by a group of equipment from Aspen Model Palette(Fig.4).MIXER was used for the reactor plenum,USER MODEL with MS Excel subroutine were used to calculate reaction kinetics and product composition,FLASH 2 enabled separation of the ebullated bed effluent into gas and liquid phase,SPLITTER was used to split liquid into reactor effluent and recalculating part.A 100%separationefficiencywasassumedfortheliquidandgasphases.Gasandliquid were mixed together as a single gas and liquid effluent.

Fig.4.EBR scheme in the Aspen Plus environment.

Table 3 List and characteristics of pseudo-components used in this work

User model in Aspen Plus allows defining custom input and output parameters.Forthiswork,theinputparameterswere:feedcomposition(distillation curve),feed rate,reactor outlet temperature,and consumption of hydrogen.A set of pseudo-components was generated(Table 3)from the distillation curves and densities of feed and product fractions.Resulting39 pseudo-componentsare represented bytheaveragenormal boiling point,gravity,molecular weight,critical temperature and critical pressure.To calculate output parameters of the model,and the compositionandparametersoftheoutputstream,asubroutinewaswritteninMS Excel with the user model providing communication between the Aspen environment and the subroutine.The Excel subroutine consisted of kinetic model equations(Eqs.1-7)and a pseudo-component distribution model.The distribution model converts lumped fractions calculated by the kinetic model into the respective pseudo-component blends.The lump mass fraction was divided by the total number of pseudocomponents belonging to the same lump.Gasoline and heavier lumps were subsequently represented as a mixture of pseudo-components in the Aspen Plus simulation environment.Hydrogen and hydrocarbon gas were represented by real components.Hydrocarbon streams were subsequently represented as a mixture of pseudo-components in Aspen Plus simulation environment.Due to high pressure,high temperature and a wide range of components in the reaction mixture,the Redlich-Kwong-Soave(RK-Soave)propertymethodwasusedforthesimulation.The RK-Soave property method[11]uses the Redlich-Kwong-Soave cubic equation of state for all thermodynamic properties except for the liquid molar volume,the API method for the liquid molar volume of pseudo-components,and the Rackett model for real components.

4.Model Verification

The model was verified by two different methods:

(1)Comparison of the measured distillation curve of blended final products from an industrial scale unit and the calculated distillation curve of effluent from the third reactor stage.

(2)Comparison of measured and simulated temperatures of the second and third reactor inlets after quenching by hydrocarbon oil and hydrogen stream.

As it is shown in Fig.5,the agreement between the experimental and model distillation curves is very good.Operational conditions are defined in Table 1—Case 2.Only in the region of very low boiling pointcomponents,thedifferencebetweenthemeasured andcalculated data is significant.

Thecomparisonofmeasuredandsimulatedtemperaturesofthesecond and third reactor inlets after quenching by hydrocarbon oil and hydrogen stream is shown in Table 4.In this table,the results for the two chosen cases are presented.Deviation ranging from 0.4 °C to 4 °C can be considered as acceptable.

5.Results and Discussion

Fig.5.Comparison of generated and measured distillation curve of reactor effluent.

Table 4 Comparison of measured and simulated reactor inlet temperatures after quenching

Table 5 Calculated gas and liquid properties for each reactor stage

The objective of the simulation was to calculate the recycle massflow for each reactor stage as well as the properties of gas and liquid phases(Table 5).Input for the simulation was:mass flows,pressure and temperature of individual inlet streams shown in Fig.2.The recycle mass flow was calculated by varying the split ratio in SPLITTER B6(Fig.4)for the liquid recycle and effluent to match the temperature in the reactor plenum.

Density and molecular weight of the gaseous phase increase from the first stage to the third with the hydrocracking conversion progress,hydrocarbon components are generated in thegaseous phase while hydrogen is consumed.Comparing the properties of the gaseous phase in Case 1 and Case 2,it can be seen that at higher reactor temperatures(Case 2),heavier hydrocarbons are evaporated to the gaseous phase.

Molecular weight of the liquid phase decreases in each subsequent reactor stage and with higher reactor temperature as the heavy liquid components crack and evaporate into the gaseous phase.On the contrary,density of the liquid phase does not change with temperature or conversion.It can be assumed that high molecular hydrocarbons do not vary in density significantly.Calculated densities of the gas and liquid phases in this work fit the values of 40 kg·m?3for gas and 703 kg·m?3for liquid reported by McKnight[12]for industrial scale residual ebullated bed hydrocracking reactor.

Fig.6.Feed rate and recycle ratio during the considered period.

Recycle ratio is represented by the ratio of calculated recycle flow and total feed to the respective reactor stage.For Case 1 and Case 2,the recycle ratio varies from 5.8 to 11.3.For detailed evaluation,daily averaged data from a 150 day time lapse(Fig.6)were submitted to simulation.All three recycle ratios showed a negative correlation to the feed rate.When the feed inlet to the reactor decreases,catalyst in the ebullated bed starts to settle.Settling catalyst is detected as a decrease in the density measured by the nuclear density meter and the control system subsequently increases the ebullating pump frequency.Ratios for the first and the second stage show very similar behavior over the considered period,with the average ratio of 8.7 for the first and 6.4 for the second stage.Average third stage recycle ratio is 14.5 and shows higher sensitivity to a feed rate change;peaks on day 32 and 139 did not appear for the first and second stage ratios(Fig.6).Much higher third stage recycle compared to the first and second ones is assumed to be caused bylower totalcatalystinventory andlower viscosityoftheliquidphaseinthethirdstageduetohydrocrackingconversion.Lower catalyst inventory is also indicated in Fig.7,where the third stage reactor has the lowest pressure drop across the ebullated bed.

Fig.7.Calculated recycle ratios versus measured pressure difference between pump discharge and suction.

Recycle ratio calculated in this work matches approximately the values reported in literature.Gauthier et al.reported the liquid recycle ratio for the bench unit is typically 40 and for industrial units it can be ten times lower due to the differences in design[13].Schweitzer and Kressmann[14]explained that due to the low reactor diameter in the bench scale unit,wall-particle friction forces are high compared to the particle-particle friction forces.Thus,a high recycle ratio(of about 15),with respect to the feed flowrate,is necessary to fluidize the bed in a smaller scale unit.For the industrial scale reactor,where the diameter is high,the particle-wall interactions become negligible compared to the particle-particle interactions and a low recycle ratio of about 3 suffices to fluidize the bed.McKnight[12]stated that recycle liquid for an industrial scale unit represents approximately 85%(recycle ratio of 5.7)of the liquid phase passing through the reactor[12].

Direct representation of the recycle ratio as a function of feed rate is shown in Fig.8.All three ratios are descending in the whole feed rate range as the total flow through the ebullated bed has to be maintained at the required value.The higher feed rate increases the liquid fraction of the inlet mixture and the density of the ebullated bed,which subsequently requires lower amount of liquid recycle.All three ratios show a steeper slope at the startup conditions with the feed rates between 90 t·h?1and 100 t·h?1.For this feed rate range,proportionally higher liquid recycle is needed to maintain the ebullation.When the unit reaches the designed throughput,the slope becomes moderate and almost constant for the second stage.

Fig.8.Feed rate and recycle ratios during the considered period.

In further investigation,the recycle ratio in each reactor stage was plotted against the respective measured pressure difference between the recycle pump discharge and suction.From Fig.7 the decreasing trend of the recycle ratio and the pump pressure difference can be clearlyseen.Suggestedexplanationisthatthepumppressuredifference is predominantly dependent on the catalyst inventory amount in ebullated bed which creates flow resistance.Before the industrial unit startup,the reactors are usually filled with approximately equivalent amount of fresh catalyst.After several years of operation and catalyst additionandwithdrawalsequences,thecatalystinventoryineachreactor stage can vary significantly.Where the pressure difference is lower,lowercatalystinventorycanbeassumedandthus,higherrecycleratiois required to maintain the required level of the ebullated bed.Higher dispersion of the recycle ratio of the third stage observed in Fig.7 and Fig.8 is caused by the more sensitive change of catalyst addition and withdrawal process over the measured period of time.It is assumed that the total catalyst inventory is lower for the third stage than for the first and second ones.Low catalyst inventory and low liquid phase viscosity in the third stage resulted in more sensitive recycle pump operation leading to more scattered data.

One of the advantages of ebullated bed reactors is their nearly isothermal regime.The recycle ratio has direct impact on the temperature gradient of the ebullated bed.Recirculation returns hot material into the reactor plenum and preheats the reactor inlet which reduces the temperature gradient(Fig.9)and shifts the conditions towards the isothermal regime.Higher temperature gradient for the first and second reactor stages is caused by fresh unsaturated feed,which supports the highly exothermic reactions.By increasing the recycling ratio,the temperature gradient across the ebullated bed decreases.In the third stage,where feed is partially saturated and recycle ratio is high,the temperature gradient ranges only from 2 to 4°C.

Fig.9.Effect of recycle on ebullated bed temperature gradient.

6.Conclusions

Using measured data and Aspen Plus simulation,a model of an ebullated bed reactor cascade for residual hydrocracking capable to calculateinternalstreams'parametershasbeendeveloped.Themodelwas verified by comparison of experimental and calculated distillation curves and of reactor inlet temperatures of the second and third reactors.Differences between measured and model data were insignificant and the model was used to investigate liquid recycling in a reactor cascade and itseffect onselected processparameters.Two specific regimes were found:in the range from minimal feed rate up to 100 t·h?1,proportionally higher recycle is needed to maintain ebullation of the catalyst bed,at feed rates above 100 t·h?1,moderate drop in the recycle ratio with the increasing feed rate was observed.The calculated recycle rates and power consumption of the ebullating pump match the design parameters except for the third stage pump.

Arelationbetweentheliquidrecyclingflowrate,pumppressuredifference and catalyst inventory has been identified.Further investigations in this area can led to the optimization of the catalyst inventory.Influence of the recycle ratio on the temperature gradient along the reactor decreases with the cascade stage,with strongest influence in the first stage.Temperature gradient in the third stage is practically not influenced by the recycle ratio.Results of this work can be used as a basis for ebullated bed reactor hydrodynamic modeling.

Nomenclature

Aipre-exponential factor,min?1

Eaiactivation energy,kJ.mol?1

f fraction index

i reaction pathway indices 1-8

kikinetic constant,min?1

R gas constant,J.mol?1.K?1

T temperature,°C

t reaction(retention)time,min

U objective function

w mass fraction

Acknowledgement

This work was supported by the Grant APVV-15-0148 provided by the Slovak Research and Development Agency.The authors would like to thank Robert ?ajdlík and Juraj Sláva from Slovnaft,a.s.for providing valuable comments.