Process Systems Enterprise Limited
email this page print this page

Fluidized-bed reactors

Simulation, modeling and optimization of a pyrolysis FBR

Fluidised bed reactors have traditionally been very difficult to model successfully.

Pyrolysis ICFG process

Fig. 1(a) – Ebara ICFG pyrolysis unit

 

Pyrolysis process schematic

Fig. 1(b) – Pyrolysis process schematic. Sand is used as a circulating medium for energy transfer.

 

ICFG pyrolysis flowsheet – click to enlarge

Fig. 2 – ICFG process flowsheet

 

Apart from the complex catalytic reaction, there is the question of the different times that different particles remain in the reactor.

There is a growing demand for detailed predictive modelling of particulate systems, both reacting and non-reacting. PSE has a strong background in this area, through modelling of crystallisation, granulation, fluidised bed reactors with circulating bed (i.e. population of recycled solid particles) and heterogeneous polymerisation (for example, population of swelling multigrain particles of polymer involving a large number of chemical species).

This example describes the modelling of an Internally Circulating Fluidised Bed Gasifier (ICFG) for pyrolysis of polymerising compounds of Cracking Heavy Oil (CHO).

[Acknowledgment: Ebara Corporation, Environmental & Energy Engineering Office; first shown at the PSE Japan User Meeting, September 2006].

The process

The process is shown schematically in Fig. 1(b), with the actual process shown in Fig 1(a).

Cracking heavy oil is mixed with steam and introduced to the Gasification Chamber (GC).

Energy is carried from the Combustion Chamber (CC) by a coarse grade of sand used as the circulating energy carrier. This causes gasification, with the pyrolysis products leaving from the top of the CG.

Sand coated in coke is cycled to the Combustion Chamber, where the coke is burned off, leaving the system as exhaust gas.

Modelling details

The model involved coupling a particle population balance, in order to account for a distribution of particle residence times, with CSTR mass and energy balances.

It also needed to take into account the fact that the rate of interaction of a particle with its immediate environment depends on particle age.

The gPROMS process flowsheet of the IFCG, with the Gasification Chamber on the left and the Combustion Chamber on the right, is shown in Fig. 2.

As can be expected, the model was relatively complex. Some of the more important aspects are described here.

Characterisation of sand particles. The state of a sand particle was characterised by:

2-D population balance equations – click to enlarge

Fig. 3 – 2-D population balance equations

 

Normalised dimensionless properties were introduced to describe these values.

A key issue was the fact that not only do the different particles in the reactor have different states (θ1 and θ2), but that these change at different rates (G1(θ1,θ2) and G2(θ1,θ2)) depending on the value of the state variables.

In order to take this into account, the population of particles was described in terms of a 2-dimensional distribution of θ1 and θ2. The resulting 2-dimensional population balance equation is shown in outline in Fig. 3.

Gas–solid coupling. This was described using an equation to calculate the rate of generation of gas species i by a sand particle at temperature θ1 and deposited coke and θ2.

It was necessary for the gas-phase mass and energy balances to take into account all particles in all states throughout the entire reactor.

Stage-wise approach

The project was performed in 5 stages:

Step 1 – Feed characterisation

Cracking Heavy Oil is a complex mixture. The feed was represented as a mixture of hydrocarbon components; however, for reaction, it was necessary to determine the CHO compositions in terms of weight percent of representative species.

This was done using parameter estimation. Predicted values were compared with the measured structural parameters, – a typical standard feed assay for Cracking Heavy Oil, comprising aromatic carbon rate, H-to-C atomic ratio, etc. – with good results.

Pyrolysis reaction schemes – click to enlarge

Fig. 4 – Pyrolysis reaction schemes

Pyrolysis reaction rate equations

Fig. 5 – Pyrolysis reaction rate equations

Yield dependence on temperature and residence time – click to enlarge

Fig. 6 – Parameter estimation: yield dependence on temperature and residence time

 

 

Yield dependence on temperature and residence time – click to enlarge

Fig. 7 – Coke deposition based on gas temperature and incoming sand

 

Yield dependence on temperature and residence time – click to enlarge

Fig. 8 – Population density of temperature and coke deposition for sand population in ICFG (3D view)

Step 2 – Formulation of reaction scheme for pyrolysis

An outline of the reaction scheme is shown in Fig. 4 (click to enlarge).

A simple Arrhenius rate expression, as shown in Fig. 5, was used. 44 parameters were estimated.

Step 3 – Estimation of kinetic parameters

The kinetic parameters were estimated using gPROMS model validation facilities, based on data from a 4 cm tubular micro-reactor rather than the full-scale fluidised bed.

This allowed accurate data to be determined in conditions that were as isothermal as possible, without the complication of extraneous effects introduced by the large-scale equipment.

The validated reaction set was thus valid for all reaction conditions, and no parameters needed to be re-estimated for the simulation of the larger reactor.

Fig. 6 shows the yield dependence on temperature and residence time for various components.

Step 4 – Modelling of ICFG unit

The modelling is described above.

Step 5 – Engineering study of reactor design

Once the model was constructed, it was used for conceptual reactor design, with a number of cases simulated to provide performance predictions. The results are summarised below.

Summary of main results:

The conceptual design runs determined values for variables such as:

In addition, it was possible to investigate the highly nonlinear effects of particle temperature distribution on the pyrolysis reaction, which was magnified with the increase in difference between the average temperature in GC and temperature of incoming sand from CC.

The investigation showed that a population balance approach is essential.

It was also noted that the yield of gas products and coke was higher than in a similar fixed-bed reactor. Once again, the population balance approach able to explain the observed differences.

Conclusions

The project showed that predictive modelling of fluidised bed reactors is possible using multi-dimensional population balances, and this is readily handled by gPROMS.

It was possible to perform kinetics identification using small-scale equipment, rather than the full-sized fluidised bed. Small-scale experiments are (a) much easier to carry out and (b) lead to accurate parameters that can be used in the full-scale design without modification.

Mixing imperfections in gas and/or solid phase handled using hybrid gPROMS multizonal/CFD models multiphase CFD (FluentŪ) for gas/solid and liquid/solid systems. This was achieved by extending the gPROMS multizonal CFD software interface.

The approach described here is suitable for systems with up to 3 to 4 solid state variables. PSE has developed a proprietary solution approach for systems with many more dimensions (for example, polymer particles with hundreds of internal states).