NoteA simple approach to improve the robustness of equation-oriented simulators: Multilinear look-up table interpolators
Introduction
Equation-oriented process simulators solve the system of equations simultaneously. Although useful when the number of recycle streams in the process is large, this approach prevents the implementation of strategies for circumventing local convergence problems (i.e., including ad hoc algorithms within the models of specific process units, the “modules” of the sequential modular simulators). According to Morton (2003), processes with awkward nonlinearity and high recycle complexity should be solved by equation-oriented simulators, provided their robustness undergoes some improvements. Nevertheless, while such generically robust solver is not available, some strategies can be developed to circumvent this bottleneck. As a general case, these strategies rely on simplifications of the models that constitute the chemical plant. However, these simplifications generally imply loss of accuracy and operational flexibility (Morton, 2003).
Here, an alternative approach is presented: using grid-based look-up tables to interpolate the results of previous offline simulations, for the process unit models whose nonlinearity was hindering the convergence of the simulations of the overall industrial plant. With this simple strategy, the accuracy of the simplified model (the interpolator) can be tuned, while still requiring low computational power (Nelles, 2001). As a proof of concept, an integrated first and second generation ethanol biorefinery was modeled, using both sugarcane and sugarcane leaves as feedstock. Two process sections, chosen for their characteristically nonlinear behavior, were used for testing the methodology: the distillation columns trains, which were simulated in steady-state, and the enzymatic hydrolysis reactor, which was simulated dynamically, with the results at the end of the batch run being considered in the global process simulations. In both cases, results obtained with this methodology were compared to the ones generated by the original phenomenological models. The look-up table interpolators were then coupled to the global process flowsheet, in order to evaluate the robustness of the approach.
Section snippets
Simulator
The simulator EMSO™ (Environment for Modeling Simulation and Optimization), used in this work, is equation-oriented. This simulator has an internal object-oriented modeling language that allows insertion of new models into its internal library. Moreover, it is possible to add plug-ins for running calculations which are not suitable for the equation-oriented approach; and new solvers may be linked as dynamic libraries as well (Soares and Secchi, 2003).
Grid-based look-up tables
Grid based look-up tables consist of a set
Results and discussion
Based on the accuracy criteria described in Section 2.2, a grid with 15 points (three inlet temperatures and five ethanol mass fractions) was used for approximating the behavior of the hydrous ethanol distillation column train. Fig. 2 presents the error distribution as a function of the two input variables (wine temperature and molar fraction of ethanol). Two output variables were selected as examples of the error distribution: the heat duty in the first column reboiler (the main energy
Conclusions
In this work, a methodology involving the use of grid-based look-up tables in equation-oriented simulators was developed. The methodology allowed the construction of simplified models that approximate nonlinear phenomenological models with a controllable accuracy, without the impairments in convergence caused by the rigorous models. As a demonstration of the capabilities of the methodology, an integrated first and second generation (1G–2G) ethanol production process (with 26,502 process
Acknowledgements
This work was financially supported by CAPES, CNPq and FAPESP (Proc. 09/13325-0 and 08/56246-0).
References (10)
- et al.
Kinetic modeling for enzymatic hydrolysis of pretreated sugarcane straw
Biochem Eng J
(2015) - et al.
Assessing the production of first and second generation bioethanol from sugarcane through the integration of global optimization and process detailed modeling
Comput Chem Eng
(2012) - et al.
A new environment for modelling, simulation and optimisation
Comput Aided Chem Eng
(2003) - et al.
Systematic methods of chemical process design
(1997) Simulation of ethanol production process from sugar and sugarcane bagasse aiming at process integration and energy production maximization
(2008)
Cited by (11)
Microbial oil and biodiesel production in an integrated sugarcane biorefinery: Techno-economic and life cycle assessment
2022, Journal of Cleaner ProductionCitation Excerpt :For the reactors (boiler, biomass pretreatment, bioreactors, etc.), global mass balances and component mass balances are considered, which take into account the consumed raw material and the formed products, based on the stoichiometric coefficients of the reactions and on pre-specified conversions relative to the limiting reagent. Detailed information about the models and assumptions can be found in previous work (Elias et al., 2021; Furlan et al., 2013, 2016; Longati et al., 2020b). The software used for simulating the process was the free “Environment for Modelling, Simulation, and Optimization”, known as EMSO, an equation-oriented, general purpose process simulator with an embedded modeling language.
Strategies to reduce the negative impact of inhibitors in biorefineries: A combined techno-economic and life cycle assessment
2022, Journal of Cleaner ProductionRetro-techno-economic-environmental analysis improves the operation efficiency of 1G-2G bioethanol and bioelectricity facilities
2021, Applied EnergyCitation Excerpt :Nevertheless, RTEA did not present targets for environmental performance. Besides, it relied on local sensitivity analysis [27], which only provides information at specific points [28] and may give misleading results for nonlinear systems. A more extensive approach, spanning the entire input factor domain would certainly be of interest [28].
Modeling and simulation of anion exchange chromatography for purification of proteins in complex mixtures
2020, Journal of Chromatography ACitation Excerpt :Non-synthetic mixtures were also utilized to investigate the ability of two different adsorption isotherms (modified Langmuir and steric mass action) to describe the process, in combination with the equilibrium dispersive model [9]. Simulations were carried out in EMSO (Environment for Modeling, Simulation and Optimization), an equation-oriented software system [17,18] that has been successfully used to simulate bio-refineries [19], fermentation processes [20], diesel production [21], and oil extraction [22]. Despite its broad applicability, EMSO has not previously been used to describe chromatographic processes.
Replacing hexane by ethanol for soybean oil extraction: Modeling, simulation, and techno-economic-environmental analysis
2020, Journal of Cleaner ProductionCitation Excerpt :The solvent vapor then proceeds to a solvent recuperation unit, see Box 1. In order to improve the robustness of the simulations of the distillation trains in the equation-oriented simulator EMSO, multilinear interpolators were used as surrogate models for these units, as described in Furlan et al. (2016). The process flowsheet for each solvent recovery system is in the Supplementary Material in Section A.1.
Virtual biorefineries: Computational tools for bridging the gap toward a low carbon economy
2020, Recent Advances in Bioconversion of Lignocellulose to Biofuels and Value Added Chemicals within the Biorefinery Concept