13.07.2015 Views

MPEC Problem Formulations in Chemical Engineering Applications

MPEC Problem Formulations in Chemical Engineering Applications

MPEC Problem Formulations in Chemical Engineering Applications

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

7 ConclusionsWith the development and widespread use of large-scale nonl<strong>in</strong>ear programm<strong>in</strong>gtools for process optimization, there has been an associated applicationof complementarity formulations to represent discrete decisions. This study <strong>in</strong>vestigatesboth the formulation and solution strategies for the MathematicalPrograms with Compementarity Constra<strong>in</strong>ts (MPCCs). Because MPCCs havedependent consta<strong>in</strong>ts at all feasible po<strong>in</strong>ts along with unbounded multipliers,special care is needed <strong>in</strong> the formulation of associated optimization problemsalong with a reliable solution algorithm. In this study, we <strong>in</strong>vestigate and summarizea number of regularization and penalty MPCC formulations and f<strong>in</strong>dthat the penalty formulation is particularly advantageous, especially when coupledwith an active set NLP solver. This conclusion is observed <strong>in</strong> a numericalcomparison with a library of MPCC test problems.Because all MPCCs can be formulated as underly<strong>in</strong>g <strong>MPEC</strong> problems thatresult from nested optimization formulations, a necessary condition for wellbehavedMPCCs is that the <strong>in</strong>ner optimization problem be convex with solutionsthat are not restricted by upper level constra<strong>in</strong>ts. This condition is used to motivatea systematic strategy for the formulation of complementarity constra<strong>in</strong>ts.This strategy is applied to a number of widely-used examples <strong>in</strong> process eng<strong>in</strong>eer<strong>in</strong>g,<strong>in</strong>clud<strong>in</strong>g flow reversals, relief valves, compressor kickbacks, controllersaturation and disappearance of phases <strong>in</strong> equilibrium stages.This comb<strong>in</strong>ation of well-posed formulations and MPCC solutions strategiesis demonstrated on two large-scale case studies with as many as 8000 discretedecisions. First, <strong>in</strong> the discretized dynamic system with switches, both thepenalty and relaxed formulations are very effective with active set and barriersolvers, respectively. Second, <strong>in</strong> deal<strong>in</strong>g with disappear<strong>in</strong>g phases <strong>in</strong> distillationoptimization models, the penalty formulation shows significant performance improvementsover the NCP formulation <strong>in</strong> a previous study. From these results,we f<strong>in</strong>d that well-posed complementarities coupled with NLP problems basedon penalty formulations are efficient and effective strategies for the solution ofMPCCs <strong>in</strong> process eng<strong>in</strong>eer<strong>in</strong>g.AcknowledgementsFund<strong>in</strong>g from the National Science Foundation under grant #CTS-0438279 isgratefully appreciated. The authors also wish to thank Dr. Steven Dirkse forfruitful discussions and help with the NLPEC package.References[1] Anitescu, M., P. Tseng, and S.J. Wright, “Elastic-mode algorithms formathematical programs with equilibrium constra<strong>in</strong>ts: global convergenceand stationarity properties,” Math. Program., 110, pp. 337-371 (2007).26

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!