- Email: [email protected]
A control strategy for extractive and reactive dividing wall columns
G Model CEP 6883 No. of Pages 6
Chemical Engineering and Processing xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Chemical Engineering and Processing: Process Intensification journal homepage: www.elsevier.com/locate/cep
A control strategy for extractive and reactive dividing wall columns Manuel Rodríguez* , Ping Zhou Li, Ismael Díaz Chemical Engineering Department, Universidad Politécnica de Madrid, 28006 Madrid, Spain
A R T I C L E I N F O
Article history: Received 7 March 2016 Received in revised form 22 September 2016 Accepted 7 October 2016 Available online xxx Keywords: Distillation Dividing wall column Extractive dividing wall column Model predictive control
A B S T R A C T
Dividing wall Columns (DWC) are being an important breakthrough in distillation technology due to the energy consumption and capital cost reduction they provide. There is an increasing amount of research papers devoted to DWCs although most of them present different applications and DWC setups like extractive DWC, azeotropic DWC or reactive DWC. These units have strong interactions in their operation and to achieve a good control is of great importance to guarantee a smooth and stable performance. Although control of standard DWCs has been presented somewhere else, in this work we present the control of an extractive and a reactive DWC. We establish the decentralized structure as well as a model predictive control and compare both approaches. ã 2016 Elsevier B.V. All rights reserved.
1. Introduction As a thermal separation method, distillation is one of the most used and therefor important separation technologies in the chemical industry. Basically, in every production process some of the chemicals go through at least one distillation column on their way from raw species to final product. Distillation is and will remain the separation method of choice in the chemical industry (about 95% of all industrial separation processes involve distillation, [17]. Despite its flexibility and widespread use, this unit operation is very energy demanding, which constitutes one important drawback. The US Dpt. of energy estimated in 2001 [15] that there are more than 40000 distillation columns in North America and that they consume about 40% of the total energy used to operate plants in the refining and bulk chemical industries, around 4.8 quadrillion BTUs and it is the responsible of 3% of the energy usage in the US [14]. Distillation resembles a heat engine producing a separation work with a rather low efficiency. Lost work (energy) in separation systems is due to irreversible processes of heat, mass transfer, and mixing, and is directly related to entropy production according to the Gouy-Stodola principle [2]. The thermodynamic efficiency of a distillation column is around 10% [8]. In order to reduce this drawback, which is important to achieve overall plant energy savings, new approaches and configurations have appeared. Conventional ternary separations progressed via thermally coupled columns such as Petlyuk configuration to a novel design
* Corresponding author. E-mail address: [email protected] (M. Rodríguez).
that integrates two distillation columns into one shell setup known today as dividing-wall column (DWC). DWC can separate three or more components in one vessel using a single condenser and reboiler. The DWC concept is a major breakthrough in distillation technology, as both energy consumption and capital cost can be reduced. In fact, using dividing wall columns can save up to 30% in the capital invested and up to 40% in the energy costs particularly for close boiling species [3,9]. Several companies like Montz, BASF or AzkoNobel are actively researching in this area [1,6]. However, the control of a dividing wall column is more difficult than the control of a conventional schema with two columns for the separation of ternary mixtures because there is more interaction among control loops. The distrust on DWC controllability and flexibility is mainly due to the complex design of the control strategy. Part of the complexity comes from DWCs having more degrees of freedom (DOF) compared to conventional distillation columns. This entails a complex design, but also presents extended optimization capabilities. If three product specifications are taken into account, DWC has 7 DOF: distillate and bottoms flowrate, reflux ratio, reboiler duty, sidestream flowrate and vapour and liquid internal split ratios. 5 DOF are used to stabilize 2 levels and 3 compositions while the remaining 2 DOF are used for optimization purposes. Traditionally, liquid split ratio (aL = LP/LM) and vapour split ratio (aV = VP/VM) are optimization variables. Vapour split ratio is usually fixed during the design stage because it is given by the pressure drop across both sides of the wall, which in turn depends on the stages type and geometry. The liquid split ratio is used as a control variable during operation by manipulating the flowrates leaving the bottom tray of the rectifying section. The optimal de-sign is given by the number of stages in the different sections of the DWC. The number of stages at both sides of the wall
http://dx.doi.org/10.1016/j.cep.2016.10.004 0255-2701/ã 2016 Elsevier B.V. All rights reserved.
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
2
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
Fig. 1. Methodology to generate the model for the MPC.
is usually the same but approaches with different number of plates have been reported. Olujic et al. presents a review on the different design approaches of DWC [10]. Maintaining product specifications while rejecting disturbances and loop interaction are the key concerns together with achieving significant energy savings. Some recent work has been presented extending DWC to other distillation setups, like extractive DWC and reactive DWC with important industrial applications like using extractive DWC for bioethanol dehydration [4,6,7]. Some work has been done related to have a good control structure for conventional DWC [5,16] or their control using multivariable predicive control [11,12] but there are very few works addressing the control of reactive and extractive DWC. In this work we present the control of extractive and reactive dividing wall columns using a decentralized approach as well as using the multivariable predictive controller approach. The remaining of the paper is organised as follows. Section two describes the methodology used to generate a model for the model predictive controller (MPC). Section three simulates and designs the decentralized control structure and the MPC for an extractive DWC. Section four presents the reactive DWC case study along with its simulation and model predictive control. Finally, section five draws conclusions and discusses the obtained results. 2. Methodology The methodology explained in this section has been applied to the extractive and reactive dividing wall columns case studies. The first step is to build a steady state simulation (in this case Aspen Plus was used for this purpose). The DWC column cannot be modelled as such using the commercial simulator so before constructing the model a thermodynamically equivalent system has to be developed. The steady state model of the thermodynamically equivalent configuration is used to size the equipments. The second step is to create the dynamic simulation (Aspen Dynamics have been used for this task). The dynamic model is automatically generated from the steady state model providing the necessary dynamic parameters for the different equipments.
Once the dynamic model is available the designed decentralized control is implemented in the dynamic simulator and its performance is recorded. The dynamic model as it is available in Aspen Dynamics is not suitable for MPC so a transformation is needed. In this case an existing API, called Control Design Interface, provided by Aspen, has been applied. In the Control Design Interface, the manipulated, controlled and disturbance variables are indicated, with that information the API generates the matrices of the linearized state space model using the rigorous dynamic model. This linearized model is the one to be used in Matlab for the MPC. The generated state space model has previously been scaled in order to avoid very different magnitudes and then implemented in Matlab. Before using the generated model for the MPC a validation has been performed. In order to validate the model, a Simulink wrapper (AMSimulation block) of the original dynamic model has been created. The linear model is validated against the rigorous model embedded in the Simulink environment (which means that the nonlinear dynamic model is being called from Simulink). Fig. 1 shows the followed procedure. The final step is to design the MPC (in Simulink) using the state space model. Fig. 2 shows the final structure. To account for the performance of model predictive controller, it is connected to the rigorous model (through the Simulink wrapper AMSimulation block). The Pct2Eng and Eng2Pct blocks are unit converters from engineering to percentage, this is necessary as the MPC is scaled. The MPC parameters are tuned (control and prediction horizons, sampling time, weighting factors for the controlled variables) and the final MPC is used to evaluate its performance under different disturbances. 3. Extractive dividing wall columns case study The case study is based on the paper by Kiss and Ignat [7] where they present the use of an extractive dividing-wall column for bioethanol dehydration. The ethanol dehydration and concentration is achieved using ethylene glycol as the extracting agent. The
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
3
Fig. 2. MPC structure in Simulink: up the rigorous embedded dynamic model, down the model predictive controller.
feed is a mixture with 10%w ethanol, the total mass flowrate 12500 kg/h. The amount of solvent used is 20783 kg/h. Fig. 3 shows the extractive DWC setup as well as its thermodynamically equivalent configuration with conventional columns, in this case three conventional columns are needed, one prefractionator (left side of the wall of the DWC) and two columns for the right part of the wall and the sections above and below it.
This equivalent configuration is the one that has been implemented in Aspen Plus and then in Aspen Dynamics. The EDWC has 42 stages being the number of stages of the prefractionator 17 The wall position is from stage 18 to 34. Solvent is fed in stage 4 and the side stream withdrawal is in stage 17. In this particular configuration the liquid back from the upper part of the wall only goes to the right part of the wall (liquid split ratio 0:1),
Fig. 3. Extractive DWC and its thermodynamically equivalent configuration along with the implemented for control.
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
4
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
Fig. 4. Decentralized control: up left ethanol composition, down left water composition, up right pressure, down right level.
liquid to the left part comes from the solvent. Vapour split ratio is set to 0.4: 0.6. This last configuration has been simulated with Aspen Plus and has been used to design the decentralized control and to generate the state space model to be used for the Model Predictive Control. NRTL thermodynamic package has been used along the simulation. The decentralized control implemented is also shown in the Fig. 3. The selected controlled variables in the simulation are: inventories (levels and pressures) and qualities (compositions). It is important to notice that there are some additional controls shown in the diagram that are not present in the actual extractive DWC but they are necessary for the simulation (for example the pressure control in the down column or the level control in the prefractioning column). Fig. 4 shows the results of applying the decentralized control. Water and ethanol compositions and column pressure and bottoms level are shown. The obtained response corresponds to a 2.5% disturbance in the feed. Although the disturbances are compensated, the system exhibits an oscillatory response, there is a strong interaction between control loops which makes decentralized control very difficult. A lot of effort has to be put into tuning parameters to improve the response and make it less oscillatory. Results for the MPC configuration are shown in Fig. 5. The same 2.5%feed flow disturbance has been applied. The figure includes the ethylene glycol composition which is controlled by inference with the temperature of a sensitive tray. The control achieves a smooth response and with a good performance. Higher disturbances have been tested with a positive response by the MPC controller. The controller includes constraints for the manipulated and controlled variables in order to avoid great step changes for the manipulated variables or unfeasible situations that cannot appear in an actual plant (like compositions greater than 1,
levels above 100% and so on). The prediction horizon is set to 20 and the control horizon to 4, the sampling time is 2 min (which makes a prediction horizon of 40 min). 4. Reactive dividing wall columns case study The case study is based on the paper by Sander et al. [13] where they resent the use of a reactive dividing-wall column for the hydrolysis of methyl acetate to produce methanol and acetic acid. This system has been widely studied in the literature to illustrate the benefits of reactive distillation columns as it reduces drastically the number of columns regarding the original process. A reactive dividing wall column was studied in Sander’s work as an improvement to the conventional reactive distillation configuration. In this case Fig. 6 shows the reactive DWC setup as well as its thermodynamically equivalent con-figuration with conventional columns (in this particular case just two columns are needed to set up the simulation). The reactive DWC has 63 stages, the wall starts in stage 10 and it continues until stage 52. The left part of the wall is simulated as a reactive column being the reactive zone between stages 6 and 17. Water is fed pure and methyl acetate is fed with 19%w of methanol. Water is set to have a mole ratio of three to methyl acetate. The model predictive controller has been developed and implemented for this system following the described methodology. The prediction horizon is set to 20 and the control horizon to 4, the sampling time is 2 min. In Fig. 7 the controller response is shown. Left figure represents the composition control for the three outputs of the RDWC, right figure shows the pressure and level control (up and down figures respectively). As it can be observed the applied 5% disturbance in the feed are handled without any difficulties by the MPC. The controller has shown a more robust
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
5
Fig. 5. MPC control: up left pressure, down left water and ethanol composition, up right level, down right ethylene glycol composition (temperature inference).
Fig. 6. Reactive DWC and its thermodynamically equivalent configuration implemented for control design.
behaviour than the obtained in the case of the extractive DWC handling higher disturbances without problems. 5. Conclusions Dividing-wall columns are a promising alternative for some processes but these are complex, very integrated units with strong interactions. Extractive and reactive dividing wall columns are still more complex operations than conventional dividing wall columns
and its control is a difficult task. In the case of the extractive dividing wall columns the decentralized as well as the multivariable predictive control have been implemented. The decentralized control shows a very oscillatory response as the result of the interactions being difficult to control, on the other hand the MPC shows a smooth response and good control being able to handle larger disturbances than the decentralized option. In the case of the reactive dividing wall column just the MPC results are presented. It also shows a good control performance. It behaves even better than
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
6
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
Fig. 7. MPC results for a 5% feed disturbance: left figure shows composition control, up right pressure control and down right level control.
the MPC of the extractive DWC, handling better larger disturbances. This work is being completed with the analysis of azeotropic dividing wall columns and it will be presented somewhere in a near future along with the MPC for DWC, E-DWC and R-DWC. References [1] C. Bravo-Bravo, G. Segovia-Hernandez, Extractive dividing-wall column: design and optimization, Ind. Eng. Chem. Res. 49 (2010) 3672–3688. [2] Y. Demirel, Thermodynamic analysis of separation systems, Sep. Sci. Technol. 39 (16) (2010) 3897–3942. [3] E.Y. Kenig, O. Yildirim, A.A. Kiss, Dividing wall columns in chemical process industry: a review on current activities, Sep. Purif. Technol. 80 (2011) 403. [4] A. Kiss, J.J. Pragt, C.J.G. van Strien, Reactive dividing-wall columns-how to get more with less resources? Chem. Eng. Commun. 196 (11) (2009) 1366–1374. [5] A. Kiss, C. Sorin Bildea, A control perspective on process intensi-fication in dividing-wall columns, Chem. Eng. Process. 50 (2011) 281–292. [6] A. Kiss, D. Suszwalak, Enhanced bioethanol dehydration by extractive and azeotropic distillation in dividing-wall columns, Sep. Purif. Technol. 86 (2012) 70–78. [7] A. Kiss, R.M. Ignat, Innovative single step bioethanol dehydration in an extractive dividing-wall column, Sep. Purif. Technol. 98 (2012) 290–297.
[8] H.Z. Kister, Distillation Design, McGraw-Hill, Alhambra, California, 1992. [9] M. Mascia, F. Ferrara, A. Vacca, G. Tola, M. Errico, Design of heat integrated distillation systems for a light ends separation plant, Appl. Therm. Eng. 27 (2007) 1205–1211. [10] Z. Olujic, Dividing wall columns? A breakthrough towards sustainable distilling, Chem. Eng. Process. 49 (2010) 559. [11] R. Rewagad, A. Kiss, Dynamic optimization of a dividing-wall column using model predictive control, Chem. Eng. Sci. 68 (2012) 132–142. [12] M. Rodriguez, J.A. Chinea-Herranz, Decentralized control and iden-tified model predictive control of divided wall columns, J. Process Control 22 (2012) 1582–1592. [13] S. Sander, C. Flisch, E. Geissler, H. Schoenmakers, O. Ryll, H. Hasse, Methyl acetate hydrolysis in a reactive divided wall column, Chem. Eng. Res. Des. 85 (1) (2007) 149–154. [14] G. Soave, J.A. Feliu, Saving energy in distillation towers by feed splitting, Appl. Therm. Eng. 22 (2002) 889. [15] U.S. Dpt of Energy, Office of Energy Efficiency and Renewable Energy, 2001, Distillation column modelling tools, DOE, Washington DC. [16] R. Van Diggelen, W. Heemink, Comparison of control strategies for dividingwall columns, Ind. Eng. Chem. Res. 49 (2010) 288–307. [17] P. Wankat, Separation process engineering, Pearson Educ. (2006).
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
Chemical Engineering and Processing xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Chemical Engineering and Processing: Process Intensification journal homepage: www.elsevier.com/locate/cep
A control strategy for extractive and reactive dividing wall columns Manuel Rodríguez* , Ping Zhou Li, Ismael Díaz Chemical Engineering Department, Universidad Politécnica de Madrid, 28006 Madrid, Spain
A R T I C L E I N F O
Article history: Received 7 March 2016 Received in revised form 22 September 2016 Accepted 7 October 2016 Available online xxx Keywords: Distillation Dividing wall column Extractive dividing wall column Model predictive control
A B S T R A C T
Dividing wall Columns (DWC) are being an important breakthrough in distillation technology due to the energy consumption and capital cost reduction they provide. There is an increasing amount of research papers devoted to DWCs although most of them present different applications and DWC setups like extractive DWC, azeotropic DWC or reactive DWC. These units have strong interactions in their operation and to achieve a good control is of great importance to guarantee a smooth and stable performance. Although control of standard DWCs has been presented somewhere else, in this work we present the control of an extractive and a reactive DWC. We establish the decentralized structure as well as a model predictive control and compare both approaches. ã 2016 Elsevier B.V. All rights reserved.
1. Introduction As a thermal separation method, distillation is one of the most used and therefor important separation technologies in the chemical industry. Basically, in every production process some of the chemicals go through at least one distillation column on their way from raw species to final product. Distillation is and will remain the separation method of choice in the chemical industry (about 95% of all industrial separation processes involve distillation, [17]. Despite its flexibility and widespread use, this unit operation is very energy demanding, which constitutes one important drawback. The US Dpt. of energy estimated in 2001 [15] that there are more than 40000 distillation columns in North America and that they consume about 40% of the total energy used to operate plants in the refining and bulk chemical industries, around 4.8 quadrillion BTUs and it is the responsible of 3% of the energy usage in the US [14]. Distillation resembles a heat engine producing a separation work with a rather low efficiency. Lost work (energy) in separation systems is due to irreversible processes of heat, mass transfer, and mixing, and is directly related to entropy production according to the Gouy-Stodola principle [2]. The thermodynamic efficiency of a distillation column is around 10% [8]. In order to reduce this drawback, which is important to achieve overall plant energy savings, new approaches and configurations have appeared. Conventional ternary separations progressed via thermally coupled columns such as Petlyuk configuration to a novel design
* Corresponding author. E-mail address: [email protected] (M. Rodríguez).
that integrates two distillation columns into one shell setup known today as dividing-wall column (DWC). DWC can separate three or more components in one vessel using a single condenser and reboiler. The DWC concept is a major breakthrough in distillation technology, as both energy consumption and capital cost can be reduced. In fact, using dividing wall columns can save up to 30% in the capital invested and up to 40% in the energy costs particularly for close boiling species [3,9]. Several companies like Montz, BASF or AzkoNobel are actively researching in this area [1,6]. However, the control of a dividing wall column is more difficult than the control of a conventional schema with two columns for the separation of ternary mixtures because there is more interaction among control loops. The distrust on DWC controllability and flexibility is mainly due to the complex design of the control strategy. Part of the complexity comes from DWCs having more degrees of freedom (DOF) compared to conventional distillation columns. This entails a complex design, but also presents extended optimization capabilities. If three product specifications are taken into account, DWC has 7 DOF: distillate and bottoms flowrate, reflux ratio, reboiler duty, sidestream flowrate and vapour and liquid internal split ratios. 5 DOF are used to stabilize 2 levels and 3 compositions while the remaining 2 DOF are used for optimization purposes. Traditionally, liquid split ratio (aL = LP/LM) and vapour split ratio (aV = VP/VM) are optimization variables. Vapour split ratio is usually fixed during the design stage because it is given by the pressure drop across both sides of the wall, which in turn depends on the stages type and geometry. The liquid split ratio is used as a control variable during operation by manipulating the flowrates leaving the bottom tray of the rectifying section. The optimal de-sign is given by the number of stages in the different sections of the DWC. The number of stages at both sides of the wall
http://dx.doi.org/10.1016/j.cep.2016.10.004 0255-2701/ã 2016 Elsevier B.V. All rights reserved.
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
2
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
Fig. 1. Methodology to generate the model for the MPC.
is usually the same but approaches with different number of plates have been reported. Olujic et al. presents a review on the different design approaches of DWC [10]. Maintaining product specifications while rejecting disturbances and loop interaction are the key concerns together with achieving significant energy savings. Some recent work has been presented extending DWC to other distillation setups, like extractive DWC and reactive DWC with important industrial applications like using extractive DWC for bioethanol dehydration [4,6,7]. Some work has been done related to have a good control structure for conventional DWC [5,16] or their control using multivariable predicive control [11,12] but there are very few works addressing the control of reactive and extractive DWC. In this work we present the control of extractive and reactive dividing wall columns using a decentralized approach as well as using the multivariable predictive controller approach. The remaining of the paper is organised as follows. Section two describes the methodology used to generate a model for the model predictive controller (MPC). Section three simulates and designs the decentralized control structure and the MPC for an extractive DWC. Section four presents the reactive DWC case study along with its simulation and model predictive control. Finally, section five draws conclusions and discusses the obtained results. 2. Methodology The methodology explained in this section has been applied to the extractive and reactive dividing wall columns case studies. The first step is to build a steady state simulation (in this case Aspen Plus was used for this purpose). The DWC column cannot be modelled as such using the commercial simulator so before constructing the model a thermodynamically equivalent system has to be developed. The steady state model of the thermodynamically equivalent configuration is used to size the equipments. The second step is to create the dynamic simulation (Aspen Dynamics have been used for this task). The dynamic model is automatically generated from the steady state model providing the necessary dynamic parameters for the different equipments.
Once the dynamic model is available the designed decentralized control is implemented in the dynamic simulator and its performance is recorded. The dynamic model as it is available in Aspen Dynamics is not suitable for MPC so a transformation is needed. In this case an existing API, called Control Design Interface, provided by Aspen, has been applied. In the Control Design Interface, the manipulated, controlled and disturbance variables are indicated, with that information the API generates the matrices of the linearized state space model using the rigorous dynamic model. This linearized model is the one to be used in Matlab for the MPC. The generated state space model has previously been scaled in order to avoid very different magnitudes and then implemented in Matlab. Before using the generated model for the MPC a validation has been performed. In order to validate the model, a Simulink wrapper (AMSimulation block) of the original dynamic model has been created. The linear model is validated against the rigorous model embedded in the Simulink environment (which means that the nonlinear dynamic model is being called from Simulink). Fig. 1 shows the followed procedure. The final step is to design the MPC (in Simulink) using the state space model. Fig. 2 shows the final structure. To account for the performance of model predictive controller, it is connected to the rigorous model (through the Simulink wrapper AMSimulation block). The Pct2Eng and Eng2Pct blocks are unit converters from engineering to percentage, this is necessary as the MPC is scaled. The MPC parameters are tuned (control and prediction horizons, sampling time, weighting factors for the controlled variables) and the final MPC is used to evaluate its performance under different disturbances. 3. Extractive dividing wall columns case study The case study is based on the paper by Kiss and Ignat [7] where they present the use of an extractive dividing-wall column for bioethanol dehydration. The ethanol dehydration and concentration is achieved using ethylene glycol as the extracting agent. The
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
3
Fig. 2. MPC structure in Simulink: up the rigorous embedded dynamic model, down the model predictive controller.
feed is a mixture with 10%w ethanol, the total mass flowrate 12500 kg/h. The amount of solvent used is 20783 kg/h. Fig. 3 shows the extractive DWC setup as well as its thermodynamically equivalent configuration with conventional columns, in this case three conventional columns are needed, one prefractionator (left side of the wall of the DWC) and two columns for the right part of the wall and the sections above and below it.
This equivalent configuration is the one that has been implemented in Aspen Plus and then in Aspen Dynamics. The EDWC has 42 stages being the number of stages of the prefractionator 17 The wall position is from stage 18 to 34. Solvent is fed in stage 4 and the side stream withdrawal is in stage 17. In this particular configuration the liquid back from the upper part of the wall only goes to the right part of the wall (liquid split ratio 0:1),
Fig. 3. Extractive DWC and its thermodynamically equivalent configuration along with the implemented for control.
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
4
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
Fig. 4. Decentralized control: up left ethanol composition, down left water composition, up right pressure, down right level.
liquid to the left part comes from the solvent. Vapour split ratio is set to 0.4: 0.6. This last configuration has been simulated with Aspen Plus and has been used to design the decentralized control and to generate the state space model to be used for the Model Predictive Control. NRTL thermodynamic package has been used along the simulation. The decentralized control implemented is also shown in the Fig. 3. The selected controlled variables in the simulation are: inventories (levels and pressures) and qualities (compositions). It is important to notice that there are some additional controls shown in the diagram that are not present in the actual extractive DWC but they are necessary for the simulation (for example the pressure control in the down column or the level control in the prefractioning column). Fig. 4 shows the results of applying the decentralized control. Water and ethanol compositions and column pressure and bottoms level are shown. The obtained response corresponds to a 2.5% disturbance in the feed. Although the disturbances are compensated, the system exhibits an oscillatory response, there is a strong interaction between control loops which makes decentralized control very difficult. A lot of effort has to be put into tuning parameters to improve the response and make it less oscillatory. Results for the MPC configuration are shown in Fig. 5. The same 2.5%feed flow disturbance has been applied. The figure includes the ethylene glycol composition which is controlled by inference with the temperature of a sensitive tray. The control achieves a smooth response and with a good performance. Higher disturbances have been tested with a positive response by the MPC controller. The controller includes constraints for the manipulated and controlled variables in order to avoid great step changes for the manipulated variables or unfeasible situations that cannot appear in an actual plant (like compositions greater than 1,
levels above 100% and so on). The prediction horizon is set to 20 and the control horizon to 4, the sampling time is 2 min (which makes a prediction horizon of 40 min). 4. Reactive dividing wall columns case study The case study is based on the paper by Sander et al. [13] where they resent the use of a reactive dividing-wall column for the hydrolysis of methyl acetate to produce methanol and acetic acid. This system has been widely studied in the literature to illustrate the benefits of reactive distillation columns as it reduces drastically the number of columns regarding the original process. A reactive dividing wall column was studied in Sander’s work as an improvement to the conventional reactive distillation configuration. In this case Fig. 6 shows the reactive DWC setup as well as its thermodynamically equivalent con-figuration with conventional columns (in this particular case just two columns are needed to set up the simulation). The reactive DWC has 63 stages, the wall starts in stage 10 and it continues until stage 52. The left part of the wall is simulated as a reactive column being the reactive zone between stages 6 and 17. Water is fed pure and methyl acetate is fed with 19%w of methanol. Water is set to have a mole ratio of three to methyl acetate. The model predictive controller has been developed and implemented for this system following the described methodology. The prediction horizon is set to 20 and the control horizon to 4, the sampling time is 2 min. In Fig. 7 the controller response is shown. Left figure represents the composition control for the three outputs of the RDWC, right figure shows the pressure and level control (up and down figures respectively). As it can be observed the applied 5% disturbance in the feed are handled without any difficulties by the MPC. The controller has shown a more robust
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
5
Fig. 5. MPC control: up left pressure, down left water and ethanol composition, up right level, down right ethylene glycol composition (temperature inference).
Fig. 6. Reactive DWC and its thermodynamically equivalent configuration implemented for control design.
behaviour than the obtained in the case of the extractive DWC handling higher disturbances without problems. 5. Conclusions Dividing-wall columns are a promising alternative for some processes but these are complex, very integrated units with strong interactions. Extractive and reactive dividing wall columns are still more complex operations than conventional dividing wall columns
and its control is a difficult task. In the case of the extractive dividing wall columns the decentralized as well as the multivariable predictive control have been implemented. The decentralized control shows a very oscillatory response as the result of the interactions being difficult to control, on the other hand the MPC shows a smooth response and good control being able to handle larger disturbances than the decentralized option. In the case of the reactive dividing wall column just the MPC results are presented. It also shows a good control performance. It behaves even better than
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004
G Model CEP 6883 No. of Pages 6
6
M. Rodríguez et al. / Chemical Engineering and Processing xxx (2016) xxx–xxx
Fig. 7. MPC results for a 5% feed disturbance: left figure shows composition control, up right pressure control and down right level control.
the MPC of the extractive DWC, handling better larger disturbances. This work is being completed with the analysis of azeotropic dividing wall columns and it will be presented somewhere in a near future along with the MPC for DWC, E-DWC and R-DWC. References [1] C. Bravo-Bravo, G. Segovia-Hernandez, Extractive dividing-wall column: design and optimization, Ind. Eng. Chem. Res. 49 (2010) 3672–3688. [2] Y. Demirel, Thermodynamic analysis of separation systems, Sep. Sci. Technol. 39 (16) (2010) 3897–3942. [3] E.Y. Kenig, O. Yildirim, A.A. Kiss, Dividing wall columns in chemical process industry: a review on current activities, Sep. Purif. Technol. 80 (2011) 403. [4] A. Kiss, J.J. Pragt, C.J.G. van Strien, Reactive dividing-wall columns-how to get more with less resources? Chem. Eng. Commun. 196 (11) (2009) 1366–1374. [5] A. Kiss, C. Sorin Bildea, A control perspective on process intensi-fication in dividing-wall columns, Chem. Eng. Process. 50 (2011) 281–292. [6] A. Kiss, D. Suszwalak, Enhanced bioethanol dehydration by extractive and azeotropic distillation in dividing-wall columns, Sep. Purif. Technol. 86 (2012) 70–78. [7] A. Kiss, R.M. Ignat, Innovative single step bioethanol dehydration in an extractive dividing-wall column, Sep. Purif. Technol. 98 (2012) 290–297.
[8] H.Z. Kister, Distillation Design, McGraw-Hill, Alhambra, California, 1992. [9] M. Mascia, F. Ferrara, A. Vacca, G. Tola, M. Errico, Design of heat integrated distillation systems for a light ends separation plant, Appl. Therm. Eng. 27 (2007) 1205–1211. [10] Z. Olujic, Dividing wall columns? A breakthrough towards sustainable distilling, Chem. Eng. Process. 49 (2010) 559. [11] R. Rewagad, A. Kiss, Dynamic optimization of a dividing-wall column using model predictive control, Chem. Eng. Sci. 68 (2012) 132–142. [12] M. Rodriguez, J.A. Chinea-Herranz, Decentralized control and iden-tified model predictive control of divided wall columns, J. Process Control 22 (2012) 1582–1592. [13] S. Sander, C. Flisch, E. Geissler, H. Schoenmakers, O. Ryll, H. Hasse, Methyl acetate hydrolysis in a reactive divided wall column, Chem. Eng. Res. Des. 85 (1) (2007) 149–154. [14] G. Soave, J.A. Feliu, Saving energy in distillation towers by feed splitting, Appl. Therm. Eng. 22 (2002) 889. [15] U.S. Dpt of Energy, Office of Energy Efficiency and Renewable Energy, 2001, Distillation column modelling tools, DOE, Washington DC. [16] R. Van Diggelen, W. Heemink, Comparison of control strategies for dividingwall columns, Ind. Eng. Chem. Res. 49 (2010) 288–307. [17] P. Wankat, Separation process engineering, Pearson Educ. (2006).
Please cite this article in press as: M. Rodríguez, et al., A control strategy for extractive and reactive dividing wall columns, Chem. Eng. Process. (2016), http://dx.doi.org/10.1016/j.cep.2016.10.004