(578f) Data-Driven Strategies for Optimization of Integrated Chemical Plants | AIChE

(578f) Data-Driven Strategies for Optimization of Integrated Chemical Plants

Authors 

Ma, K. - Presenter, Carnegie Mellon University
Sahinidis, N., Carnegie Mellon University
Rajagopalan, S., Dow Inc.
Amaran, S., The Dow Chemical Company
Bury, S. J., Dow Inc.
A chemical site usually consists of a dozen chemical plants with complex material flow interactions and heat integrations. In current practice, operations of each plant have been extensively studied and optimized using Real-Time Optimization technologies individually. However, limited effort takes the interconnected streams into account and makes decisions from a site-level perspective. Operation optimization over large-scale integrated chemical plants is inherently a difficult problem for a system of such size and complexity. As solving detailed first-principles optimization models is often not suitable for enterprise-wide optimization, chemical process simulators are playing an increasingly important role in the chemical industry. This emerging trend is a challenge for classical optimization techniques which rely on derivatives, that are difficult to calculate or unavailable from process simulators.

In this work, we propose two novel methodologies to exploit the available Aspen flowsheets and plant data. In the first approach, we develop and solve a data-driven optimization model. The impact of the level of detail of the model is investigated. In the second approach, we apply derivative-free optimization (DFO) with a novel decomposition framework. Our proposed framework significantly extends the scope of current DFO solvers to larger-scale problems. Both methodologies are studied and compared over the dimethyl ether/diethyl ether production case study.