Oligomerization Using Reactive Distillation, Design and Model Predictive Control
Oligomerization Using Reactive Distillation, Design and Model Predictive Control
)
PROCEEDINGS OF THE 33rd European Symposium on Computer Aided Process Engineering
(ESCAPE33), June 18-21, 2023, Athens, Greece
© 2023 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/B978-0-443-15274-0.50244-4
1. Introduction
Energy has an utmost important role in our society. Energy is used to produce almost all
the necessary commodities such as food, polymers, fertilizers. This energy employed to
produce these commodities has several forms such as electricity, natural gas, liquid fuels,
among others. In this sense, one sector that is essential for our society and which is
characterized by an intensive use of energy is the transport sector. This sector consumes
around 21% of the total energy demand annually, and its energy consumption is expected
to increase up to 130% by 2050 which also represents an increase in CO2 emissions.
Among all the transport means, the aviation sector is expected to grow the fastest,
doubling its energy consumptions and CO2 emission in the next 20 years. For this reason,
the aviation industry has proposed to reduce its CO2 emissions in order to achieve a more
sustainable aviation industry. The use of alternative biofuels for replacing oil-derived
fuels is considered by the International Air Transport Association (AITA) as the most
interesting alternative to reduce emissions (Rivas-Interian et al., 2022). These biofuels
are commonly called biojet-fuels and can be produced from biomass. Nowadays, there
1532 G. Contreras-Zarazúa et al.
The distillation column was designed and simulated in Aspen Plus using the RADFRAC
module. The selection of thermodynamic model for the simulation was performed
according to Carlson´s algorithm (Carlson, 1996). Consequently, the Soave-Redlich-
Kwong (SRK) was used as thermodynamic model to simulate the system, owing to the
mixture contains linear hydrocarbons and the pressure required for the oligomerization is
around 10-40bar. The design of the reactive distillation was performed using a sensitivity
analysis in order to achieve a suitable hydrocarbon distribution and maintain low energy
consumption. It is important to highlight, that according to Li et al., (2016), a suitable
hydrocarbon distribution corresponds to a hydrocarbon mixture in the range of C10 to
C16, being the C12 the most abundant compounds in the blend with the aim of ensuring
the suitable physiochemical properties. Finally, in order to determine the improvements
of intensification, the total annual costs of both the conventional and intensified
oligomerization process were calculated using the Guthrie´s method and data reported by
Rivas-Interian et al., 2022.
The integral absolute error (IAE) was chosen as metric to quantify and compare the
closed-loop performance of the PI and MPC controllers. The tunning of controllers were
1534 G. Contreras-Zarazúa et al.
carried out using the minimization of IAE method, to achieve this, disturbances of -1%
were performed on the manipulate variables of distillation column. Mathematically, the
IAE can be defined as follows:
∞
Integral of Absulote Error=
(IAE) ∫ y(t) − y
0 sp dt (1)
Where ysp is the value of the set point for the control variables, whereas y(t) is the value
of control variable at time t. It is important to mention, that the MPC controller uses a
state space (SS) model as predictor model, this state space model was obtained by
subspace identification. The subspace identification is a common method used to obtain
a dynamic model when the process is very complex (Li et al., 2021). The rigorous
simulation of Aspen Dynamics was used as plant model. The MPC control was
implemented in a hybrid platform, which links Aspen Dynamics and the Simulink
software. On the order hand, the feedback control using PI controllers was fully
implemented in Aspen Dynamics. Disturbances of -1% on manipulate variables were
done in order to compare the performance of control systems and to tune the controllers.
4. Results
This section shows the design and control results of the catalytic column. As
aforementioned, the design of reactive distillation column was performed by a sensitivity
analysis in order to obtain a suitable olefin distribution. In this case, designs for 20, 30,
and 40 stages were studied for the column. The results of the sensitivity analysis for the
reflux ratio and reboiler duty as representative cases are shown in Figure 3. Please note,
in Figure 3 as large values of reflux ratio have a positive effect on the production of C10
and C12, while more reflux ratio has a negative effect on C14. The C12 reaches a
maximum concentration in reflux ratios close to 40, while for C10 and C14 the maximum
concentrations are reached at higher and lower reflux ratios, respectively. Based on the
results, it is concluded that a reflux ratio of 43 is the most adequate, because for this value,
there is a suitable distribution of hydrocarbons, being C12 the most abundant and C10
and C14 olefins has a similar production. In addition, it was determined that a reboiler
duty of 200kW is suitable since at this point the hydrocarbon distribution is very similar
to that reported by Li et al., (2016). The sensitivity analysis was repeated for 30 and 40
stages finding that a greater number of stages does not benefit the production of more
C10, C12 and C14. The Hydrocarbon distribution found with the best design is reported
in Figure 3., also, some important design parameters are reported in Table 1. In the case
of total annual costs (TAC), the results indicated that a conventional process has TAC
$582,204.90 $/y. On the other hand, the intensified has a TAC of $511,128.62$/yr
representing savings on costs around 13% in contrast to conventional process.
With respect to control results, the composition of the most abundant compound in domes
and bottoms were used as control variables. In this case, the butane (C4) and dodecane
(C12) are de most abundant compounds in the top and bottom of the column, respectively.
The composition of C4 was controlled with reflux ratio, while the composition of C12 is
controlled using the reboiler duty of bottoms. Once identified the control variables, the
tunning of controllers was performed. As aforementioned, the minimization of IAE
method was used for tuning the controllers, in this case the parameters that minimize the
IAE for the PI controllers at the top of column are 40 min for integral time and 20 for
gain whereas the parameters for the PI controller in the bottom of column are 11.443 for
gain and 70.99min integral time. In the case of MPC control, it is necessary only one
controller to control the composition at the top and bottoms of the column. In this case,
the parameters that minimize the IAE are a sample time of 3 min, control horizon of 10
intervals and prediction horizon of 50 intervals. Once the suitable parameters were
determined, the control performance for both PI and MPC controllers are compared using
their respective IAE values. The results of disturbing the reboiler duty as representative
case are shown in Figure 4 whereas the IAE value is shown in Table 2
C4 control loop C4 control loop C12 control loop C12 control loop
Control loop
(PI) (MPC) (PI) (MPC)
Reboiler duty
2.86E-05 0.1091 0.0010875 0.00794
disturbance
Based on the results, there is an appreciable difference between the two types of
controllers. In this case, the butene composition can be controller better using a PI
controller, having the PI controller up to 30% better performance than the MPC controller.
However, for the dodecene control loop, the performance is very similar between both
controllers. Note that the MCP control has the capability of controlling the process,
however it do not fully fits to the setpoint, being the difference small due to the tolerance
with which the objective function is solved in predictive control. However, it is concluded
that this difference is not significant and both controllers are adequate to control the
process.
5. Conclusions
In this work, a reactive distillation equipment was designed to intensify the
oligomerization stage in the ATJ process. The intensified reactive distillation column has
savings on total annuals costs of 13% in contrast to the conventional process. On the other
hand, the control studies indicated that it is possible to operate this catalytic column using
traditional control strategies, such as feedback control and other advanced control
techniques, such as model predictive control. Finally, it is concluded that both controllers
have similar performance, despite of this please note, that the PI controllers have a slight
better performance due to they returned to exactly set-point value.
References