In this section
@inbook{van:2022:10.1016/B978-0-323-85159-6.50193-7,
author = {van, de Berg D and Petsagkourakis, P and Shah, N and del, Rio-Chanona EA},
booktitle = {Computer Aided Chemical Engineering},
doi = {10.1016/B978-0-323-85159-6.50193-7},
pages = {1159--1164},
title = {Data-driven coordination of expensive black-boxes},
url = {http://dx.doi.org/10.1016/B978-0-323-85159-6.50193-7},
year = {2022}
}
TY - CHAP
AB - Coordinating decision-making capacities using optimization is a key factor in the success of chemical companies. However, this coordination is often inhibited by expensive, legally-constrained, or proprietary subproblem models. We propose two variations on how model-based (surrogate) derivative-free optimization (DFO) methods can be used to coordinate subproblems with few connecting variables. When these surrogates are convex quadratic, they can be efficiently exploited using semidefinite programming techniques. We compare the performance of these two variations with a distributed optimization solver (ADMM), a model-based, and a direct DFO solver (Py-BOBYQA and DIRECTL). This comparison is done on four variations of an economic-environmental feedstock blending optimization case study. While ADMM seems to display faster initial convergence, explorative DFO optimization solvers seem promising in escaping local minimizers, especially in lower dimensions.
AU - van,de Berg D
AU - Petsagkourakis,P
AU - Shah,N
AU - del,Rio-Chanona EA
DO - 10.1016/B978-0-323-85159-6.50193-7
EP - 1164
PY - 2022///
SP - 1159
TI - Data-driven coordination of expensive black-boxes
T1 - Computer Aided Chemical Engineering
UR - http://dx.doi.org/10.1016/B978-0-323-85159-6.50193-7
ER -