(231f) Optimal Operation of Heat Exchanger Networks

To save operational costs and reduce energy usage in industrial chemical plants, heat integration and heat recovery are some of the most important and widely applied technologies. Typically this includes networks of heat exchangers, which are used to transfer heat from a set of hot streams to a set of cold streams.

Although heat integration is generally considered extensively during plant design, its potential is often not maximized under plant operation, when the conditions are different from the nominal conditions assumed during design. Hence there is still a large potential for energy savings by optimally operating the heat exchanger networks.

A common flow configuration in practice is a feed stream which is split into two or more branches that are heated separately in heat exchangers using heat from a set of hot streams. In such cases, it is often desirable to distribute the common feed stream between the branches in such a way that the total heat transfer is maximized. If the branches are merged again, this is equivalent to distributing the stream between the branches such that the merged stream assumes the maximally possible temperature.

However, determining the optimal split is not trivial; mainly because of disturbances such as varying hot stream and utility temperatures, varying flow rates, and unknown heat transfer coefficients. One approach to determine the optimal stream split is online real-time optimization (RTO) [1,2]. This approach includes developing a mathematical model of the heat exchanger network, and using online measurements to adapt the model parameters to match the real plant. Then the adapted model is used in a numerical optimization scheme to determine the optimal flow into each branch. Although this approach can give close to optimal performance, it is expensive and complicated to implement.

An alternative approach is to determine “self-optimizing” controlled variables [3], which when controlled at constant setpoints, indirectly lead to optimal or near-optimal adjustment of the flows. Based on the ideas in [4], we present a surprisingly simple function of only temperature measurements, which when controlled to equal values for each branch, results in a stream split which maximizes the total heat transfer. Interestingly, although only temperature measurements are used, also flow disturbances, and varying heat transfer coefficients are handled in a near-optimal manner. This is very attractive for practical implementation, because temperature measurements are much cheaper and easier to obtain than flow measurements or estimates of heat transfer coefficients.

Finally, we demonstrate the application of the method on a simulation study and present generalizations for the case that the hot streams have different costs for using them.

[1] T. Lid, S. Strand, S. Skogestad, "On-line optimization of a crude unit heat exchanger network" In: Proceedings of CPC VI, AIChE Symposium Series vol. 98. pp. 403 - 407. (2002)
[2] T. E. Marlin, A. Hrymak, "Real-time operations optimization of continuous processes" In: Proceedings of CPC V, AIChE Symposium Series vol. 93. pp. 156 - 164. (1997)
[3] S. Skogestad "Plantwide control: the search for the self-optimizing control structure", J. Proc. Control, 10, 487-507 (2000).
[4] J. Jäschke and S. Skogestad, "Optimal controlled variables for polynomial systems", Journal of Process Control, 22 (1), 167-179 (2012).