(632a) Global Optimization of Real Time Operation of an Industrial Gas Network | AIChE

(632a) Global Optimization of Real Time Operation of an Industrial Gas Network


Puranik, Y. - Presenter, Carnegie Mellon University
Sahinidis, N., Carnegie Mellon University
Li, T., Air Liquide
Gopalakrishnan, A., Air Liquide
Besancon, B., Air Liquide

Air Liquide operates a number of industrial gas pipeline networks around the world. Each of them connects several air separation plants and a number of customers. In each of the plants, there can be one or more air separation units (ASU), which separates air into pure oxygen, nitrogen and argon in either gas or liquid phase. Usually, the gaseous oxygen and nitrogen supply different pipeline sections at different pressure levels after being pressurized either in the ASU or by compressors downstream of the ASU.

The operation of such a network of plants, pipelines and customers is complicated due to the rapidly fluctuating electricity prices and customer demands. To efficiently operate this network under these conditions, real-time optimization techniques are necessary. The real-time optimization model was developed by Air Liquide to describe both plant and pipeline operations through mass balance, energy balance and regression on historical data. Nonconvex functional forms are necessary to accurately describe plant characteristics. Also, certain discrete decisions must be made to choose whether certain compressors and air separation units should be turned on or off. Due to these complications, the resulting model is a mixed integer nonlinear programming problem (MINLP), with multiple local optima.

Global optimization techniques are thus essential to exploit this model and to make the best operating decisions. For use in a real-time optimization setting, this model must be solved in a reasonable time. Furthermore, the size and numerical features of the problem pose a challenge in obtaining a reliable and robust solution to the model.

We will present the collaborative effort between Carnegie Mellon University and Air Liquide in addressing the above challenges. We will present the strategies employed to tackle each of the challenges, along with computational results.