BIOMASS PYROLYSIS
IN DENSE FLUIDIZED BED
Application to the industrial case

Application to the industrial case


March 13, 2008

Contents

1 Introduction

As the preliminary study has been validated by the reference case of “Lathouwers and Bellan”, an industrial application can then be studied. In this part, EDF needs are examined and recommendations are emitted to optimize the gasifier. Indeed EDF plans to build an upgraded version of the Dual Fluidized Bed (DFB) reactor of Gussing in Austria. The shape of this reactor is presented on figure 1.


PIC

Figure 1: Geometry of the reactor


EDF launched a thesis in 2004 to develop a 1D model in order to calculate the evolution of important parameters. Strong hypothesises were made to build this model. The present work is first aimed at verifying the validity of those hypothesises thanks to a 2D simulation with the NeptuneCFD code. The major hypothesises that needs examination are that :

The present work is then aimed at examining the influence of the reactor size on the olivine fluidization and at studying the hydrodynamic behaviour of the other phases in this specific reactor. The hydrodynamics will then be modelled. The first step is to model the configuration with only one solid phase. Olivine is injected by the right upper part of the reactor and evacuated by the outflow located at the bottom of the reactor. The study of the olivine bed’s hydrodynamics is essential because the olivine is the heat provider for biomass pyrolysis.

2 The olivine bed hydrodynamics

The time average study answers to the question of our bed’s behaviour. This study provides necessary information concerning the state of the fluidized bed of olivine. For this simulation, the olivine is injected by the right side of the empty reactor and it goes out by the bottom of reactor presented on figure 1. First the simulation runs during 10 seconds to get a stabilized bed. Then the fluidization is simulated during 30 seconds to get significant average values.


PIC

Figure 2: Average volume fraction of olivine



PIC

Figure 3: Average horizontal velocity



PIC

Figure 4: Average vertical velocity


The figures 2, 3, 4 show the behavior of the olivine particles in the fluidized bed. One can see that the vertical velocity is negative on the left part of reactor whereas it is positive on the right side. Moreover, the horizontal velocity is positive at the bottom of the bed and becomes negative at the top of the bed. Consequently the olivine bed has a circulating motion which improves the mixing in the bed. This simulation then enabled to understand the behaviour of the fluidized bed and lead us to the study of the injection effects recommended by EDF.

3 The injection

Under recommendations of EDF, the stability of the fluidized bed has been studied for different injection modes. In fact EDF would like to know what would be the best way to inject the olivine into the reactor if the production capacity was doubled. As a consequence, for this simulation, the reactor size is twice larger than the previous one, the input flow has been doubled too and the output has been adjusted to have an output flow equal to the input flow. The aim of this simulation is now to know if several injections are necessary or if it is better to keep only one injection point.

Results of simulations given on figures 5, 6 show the cases with one injector on figure 6 or two injectors on figure 5. The fluidized bed has globally the same behaviour.


PIC

Figure 5: Volume fraction of olivine at t=10s for two injection



PIC

Figure 6: Volume fraction of olivine at t=10s for one injection


But the time average study illustrated by figures 7 and 8 reveals a difference between the two cases. As a matter of fact, when there are two injectors, the horizontal and vertical absolute velocities are greater than when there is just one injector. As a consequence, the use of two injectors improves the mixing.


PIC

Figure 7: Average horizontal velocity



PIC

Figure 8: Average vertical velocity


To conclude, it would be easier and cheaper to have only one combustion chamber and one injector than several ones but if there were biomass particles staying at the top of the bed, or if the fluidization needed to be improved, it would be more appropriate to have several injectors. This section closes the monodisperse study that enables to understand the olivine bed behaviour in different configurations.

4 The polydisperse hydrodynamics without mass transfer

The first step to study the polydisperse hydrodynamics is to test a case with three phases (vapor, olivine, biomass) without any reaction. The purpose of this first step is to check if the biomass particles are moving towards the top of the fluidized bed at a velocity equal to the one of the gaz. As there is no reaction, the biomass can not be injected continuously. That is the reason why there are three steps in this simulation. The reactor presented on figure 1 is modified by cutting the bottom outflow section. As the biomass can not be evacuated, the injection section is closed and we keep a constant mass of biomass in the reactor. Initially, the reactor bed is composed of the following volume fractions : 70% of vapor, 20% of olivine and 10% of biomass under the height of 1.6m. In this configuration, the bed is fluidized during 5s without any injection. Then, biomass is injected during 2s and finally one can observe the fluidization during 8s without any injection of biomass. The figure 9 presents the fields of volume fraction without mass transfer after 15s of fluidization.


PIC

Figure 9: Fields of Volume fraction at t=15s


The volume fraction of olivine and biomass were plotted on a vertical line located in the middle of the reactor as shown on 10.


PIC

Figure 10: Plotting line


The result is given by the graph presented on the figure 11.


PIC

Figure 11: Volume fraction over line at t=15s


Those results show that the biomass particles are not moving toward the top of the bed but are drifted by the olivine particles in the bed. Moreover particles are moving at a velocity equal to the one of the olivine bed. To summarize, there is a complete correllation between the motion of biomass and olivine particles. This result does not confirm the 1D hypothesis of EDF. However, as those reactions were realized without mass transfer we tried to study the hydrodynamics with mass transfer so as to get closer to the real situation.

5 The polydisperse hydrodynamic with mass transfers

The second step to study the polydisperse hydrodynamic is to test a case with three phases (vapor, olivine, biomass) with a mass transfer from the biomass phase to the vapor phase. Theoretically, referring to the validation case of “Lathouwers & Bellan”, the specific time for the reaction was 0.02s. However those simulations were realized for small diameters particles d= 0.05 cm compared to the currently studied particles which diameters is d= 4 cm. As the diameter is different, the specific reaction time does not fit with the kinetics. That is the reason why a specific time of 10s was chosen for this reaction. This simulation shows that during the reaction, biomass is reduced and produces gases. This reaction creates a gaseous flows which increases the vapor velocity and by the fluidization of the bed. The biomass particles evaporate and this gaseous flow drifts both the biomass and the olivine bed as observed on figure 12.


PIC

Figure 12: Fields of volume fraction at t=10s


6 Influence of the specific reaction time

To observe the behaviour of the bed, we take an initial reactor bed composed of 40% of olivine and 5.5% of biomass in volume on a height of 1.6m. This bed is fluidized during 10s without any injection. This simulation aims at studying the influence of the specific reaction time on the hydrodynamics in the reactor. That is the reason why two specific reaction times were tested. The reference case called ’temoin’ which has a specific reaction time of 10s was compared with a slower reaction defined by a specific time of 13s. Those specific reaction times do not refer to the real reaction but were ony chosen to prove the influence of the mass transfer on the fluidization.

Different parameters were plotted on the vertical line located in the middle of the reactor. The figures illustrate the differences of behaviour for those two reactions. The continuous line shows the local values whereas the non continuous line represent the average value calculated on the height of the line.


PIC

Figure 13: Volume fraction of biomass


The figure 13 presents the volume fraction of biomass after 10s. The distribution is not the same but there is globally more biomass in the bed for the slow reaction (blue curve) than for the other one.


PIC

Figure 14: Volume fraction of olivine


The figure 13 presents the volume fraction of olivine after 10s and shows that even if the distribution is different the average is the same.

According to those first two results, there seem to be a difference regarding the hydrodynamics for different specific reaction times. As it has already been mentionned, when looking at the temporal evolution of the bed, we noticed that the mass transfer accelerates the fluidization. It is this point that we tried to highlight by studying the velocity of the gases and the biomass in the reactor.


PIC

Figure 15: Gas velocity


One can notice on figure 15 that the gas velocity is globally more important for the quicker reaction (tau=10s) represented by the red curve. This result proves that the mass transfer influences the gas flow in the bed. As a consequence the other particles motion also increases. This point is confirmed by the study of the biomass velocity.


PIC

Figure 16: Biomass particles velocity


Indeed on figure 16, we can first notice that on the plotted line, the biomass velocity is globally positive at the bottom of the bed whereas it becomes negative in the upper side of the bed. This result shows that there is a recirculation of the biomass particles. Next, the average velocity is negative but its absolute value proves that the velocity is less important for the slow reaction (tau =13s).

This study of the specific time reaction shows that the mass transfer influences the fluidization of the bed. The results found by plotting on the line after 10s are representative of the system behaviour and were confirmed by an average time study. The different graphs illustrate that, the more the bed is reactive, the more the gas transfer is explosive and the more the solid particles are drifted. Moreover, one can see on figures 15 and 16 that the vertical velocities of the gas and the particles are different. In fact, the biomass and the olivine have globally the same velocity in the bed. However, considering the limits of NeptuneCFD and the lack of time, we were not able to improve the understanding of the hydrodynamics when mass transfer occur.

7 The limits of the software

However, this simulation does not completely fit with the real physics because the real reaction also produces a solid phase of char with different properties than the biomass phase. This former phase would have modified the hydrodynamics but were not modelled because of the lack of time.

As a matter of fact, this software easily solves two phases simulation but as the model used hardly solves the interaction between more than two solid phases, the real simulation was not realized. This limitation comes from the interaction and compacity models used to build the code. Indeed, as no other solid phase could be added, no simulation would have enabled to check if the char particles were mixed up with the bed. However, referring to the validation case studied, we could have made hypothesises concerning the transporting phases and try to implement the complex reaction with scalars in the code but we lacked time because we tried to solve the polydisperse problem.

8 Conclusion

To conclude, the study of the application provides interesting results concerning the one solid phase study. First, the simulation of the monodisperse model with the NeptuneCFD code leads us to the conclusion that the injection of olivine creates a circulating motion of the solid bed. It also showed that even if it is more expensive and not necessary, the use of two injectors tends to improve this mixing. Next, the simulation of the polydisperse model on the NeptuneCFD code provides results contradicting the 1D model hypothesis of EDF. Indeed the different simulations realized lead us to the conclusion that there is a correlation between the biomass and olivine particles : the biomass are not moving toward the top of the bed but stays in the bed at a velocity almost three times less than the gas bubbles. However those results do not mean that the 1D model hypothesis are wrong because the polydisperse model used in the NeptuneCFD code mustn’t be considered as the very truth. Indeed, the limits of the code concerning the tridisperse simulation proved that this model may encounter problems. Moveover, because of those limits and also the lack of time, we were not able to model the char and tar motion in the bed and as a consequence we could not check the hypothesises emitted by EDF. Finally, at this time, the NeptuneCFD code can not solve the simulations with more than three disperse solid phases and the results provided do not definitely contradict the 1D model but shows the necessity to compare numerical results to experiments. If we had got enough time, we could have implemented the complex reaction with scalars as realized for the validation case with NeptuneCFD. This could be a clue for the following studies.