Key Note: Heat Exchanger Networks 40 Years after the Discovery of the Pinch
- Type: Conference Presentation
- Conference Type: AIChE Spring Meeting and Global Congress on Process Safety
- Presentation Date: August 19, 2020
- Duration: 25 minutes
- Skill Level: Intermediate
- PDHs: 0.50
Optimization formulations offer tools for handling these trade-offs in a rigorous way, and they also provide a possible framework for automatic design. Both deterministic approaches (e.g. Mathematical Programming, MP) and stochastic search algorithms have been proposed to address the Heat Exchanger Network Synthesis (HENS) problem. The HENS problem involves both (i) discrete decisions such as matching of streams and sequence of units, and (ii) nonlinear relations in process models and cost equations. As a result, the problem is a so-called Mixed Integer Non-Linear Programming (MINLP) problem. The computational complexity increases exponentially with the size of the problem (e.g. the number of streams and utilities), and the non-convex nature of the non-linear relations causes gradient-based optimization algorithms to end up in a local rather than the global optimum. Stochastic search algorithms trying to mimic processes in nature (e.g. Simulated Annealing, Genetic Algorithms, Particle Swarm, etc.) try to overcome the problems related to local optima. Experiences show, however, that very long computing times are required, and there is no guarantee for global optimum. Nevertheless, good (near optimal) solutions are often obtained in reasonable times.
In order to take advantage of the best of the two schools of methods (PA and MP) while avoiding their deficiencies, hybrid methods have been developed. One such approach, SeqHENS, will be described and used in this presentation to solve an industrial size Heat Exchanger Network Synthesis problem. As the name indicates, the method is sequential in nature, and the large and complex MINLP has been broken down into more easily solved problems in the form of LP, MILP and NLP. Maximum insight from PA has been utilized to design a framework where the designer (engineer) can explore various loops where the level of heat recovery, the number of units, and the matching between hot and cold streams are varied in a systematic and efficient way.
Unfortunately, both schools of methods (PA and MP) have used rather simplified models for heat exchangers and utilities. In addition, the cost models used are very simple, and a number of issues that are important in industry have been neglected in these models. Examples of such issues are stream splitting, piping, layout, space, weight, materials of construction, pressure ratings, flow configurations, internal details for heat exchangers, as well as issues of process operation and control.
Another problem with the above-mentioned methodologies for heat recovery in industrial processes is the fact that only thermal energy (heating/cooling) is considered with temperature as the quality parameter. Mechanical energy (work and power) is not considered, and the same applies to the pressure of process streams. An important extension of HENS has been emerging over the last 10 years referred to as Work and Heat Integration, where thermodynamic principles are used to design Work and Heat Exchange Networks (WHENs) consisting of heat exchange equipment (heat exchangers, heaters and coolers) and pressure change equipment (compressors, pumps, expanders and valves) in order to improve energy efficiency while considering both heat and work. The main idea is to include heat from compression and cooling from expansion in the classical heat recovery problem. In this framework, process streams can temporarily be used as utilities or as working fluids in heat pumps and refrigeration cycles. Both PA and MP based approaches have been proposed to solve WHEN problems.
The presentation will focus on the high-level topics related to the development and applicability of the methods for heat exchanger networks (and simultaneous work and heat integration) in an industrial real-life setting, while avoiding details on mathematics and algorithms. An industrial-size problem will be used to illustrate recent developments.
|AIChE Member Credits||0.5|
|AIChE Graduate Student Members||Free|
|AIChE Undergraduate Student Members||Free|