Interpoladores multilineares como metamodelos: integração à simulação de biorrefinarias no ambiente EMSO

Abstract

The use of equation oriented simulators presents several advantages, mainly the separation between the mathematical model and the algorithm responsible for its resolution. This allows an easy implementation of new models as well as their reuse for various applications (simulation, optimization, parameter estimation, etc.). However, this same characteristic makes it impossible to use specific algorithms to solve models with difficult convergence. While generic and robust algorithms are not developed yet, some strategies that essentially consist of replacing nonlinear models with linear approximations can be used. Among the options are multilinear interpolators, which use data obtained in off-line simulations to approximate the behavior of the rigorous model. This option has several advantages, such as low computational demand and controlled accuracy. Recently, the use of interpolators coupled to equation oriented simulators was the subject of an article submitted by the group of researchers from the Bioprocess Development and Automation Laboratory (LaDABio) of the Federal University of São Carlos (FURLAN et al., 2016). However, the creation of a method to facilitate the construction of surrogate models has not yet been explored. In this context, this study aims to develop a methodology for construction of surrogate models and automation of the method, besides the evaluation of look-up tables. For this work, the methodology was applied to the train of evaporators with two effects and to the combined heat and power (CHP) stage, both sectors of a sugarcane biorefinery producing first and second generation ethanol. In both cases, the proposed methodology was able to approximate the response of the rigorous models within the accuracy specified.