Prof. Sandro Macchietto
Imperial College London, UK
Title: Optimal cleaning scheduling and flow control of heat exchanger networks under fouling: recent advances
Abstract
Fouling in heat exchangers is a major source of energy recovery inefficiencies. The unwanted deposition of material over transfer surfaces decreases heat duties, increases pressure drops, reduces energy recovery, causes lower throughputs and eventually, shut downs. An economically important case is crude oil refining, where large heat exchanger networks (pre-heat trains) at the front-end of every crude distillation unit (CDU) process essentially the entire world crude oil. Similar issues arise in the chemical, food and water industries.
Two common fouling mitigation alternatives are i) the control of flow distribution in the network (by manipulating bypasses and flow splits to parallel branches) and ii) the periodic cleaning of selected units. Flow control has a rather fast dynamic response which needs to be addressed using detailed dynamic models, while fouling is a slow dynamic process which is typically addressed using scheduling methods with highly simplified (pseudo) steady-state models. The two problems are traditionally approached independently and using distinct techniques. It is speculated there are synergistic effects arising from optimising flow control and cleaning schedules simultaneously, but it is not obvious whether this is beneficial and to which extent, hence whether a simultaneous solution, which is very challenging, is worth tackling in the first place.
Here, a simultaneous solution is proposed based on the formulation and solution of a MINLP optimization problem that considers both short and long term dynamics, with realistically complex models, over a long operating horizon (from months to a few years).
Two key aspects are addressed: i) a realistic representation of the problem, and ii) the efficient solution of the optimal cleaning scheduling and control problems. The starting point is a compact, radially distributed, but axially lumped, nonlinear dynamic heat exchanger model which accounts for fouling as a growth rate of the deposit, its composition and impact on thermal and hydraulic performance. An efficient formulation is obtained via a time horizon discretization into periods, and a continue time approach is used to model the transitions between discrete states (e.g. “operating”, “being cleaned”) in the units. This approach allows solving simultaneously the scheduling and control problems. We also describe how scheduling decisions are modelled and how this general formulation can handle simultaneous cleanings of multiple units and different types of cleanings (e.g. mechanical, chemical, which have different effectiveness in restoring thermal and hydraulic performance).
The second key aspect deals with two main numerical difficulties: i) the complexity arising from the combinatorial nature of the problem due to the large number of possible cleaning schedules over long operating horizons, and ii) the large problem size and large number of nonlinearities that arise from the use of realistic heat exchanger and fouling models. Mathematically, this translates into a large number of binary variables, complicating variables and constraint in the formulation. We propose a reformulation and relaxation of the binary variables and scheduling constraints using complementarity constraints. This allows solving the optimal cleaning scheduling problem in a reasonable computational time, with general and flexible applications of a variety of operational constraints.
A case study of industrial significance is presented that demonstrates that i) the proposed formulation and solution strategy can solve realistically large heat exchanger networks and ii) there are indeed substantial synergies and potential economic savings resulting from the simultaneous optimisation of cleaning schedules and flow rate control, over and above the optimisation of either aspect individually. This is well explained in terms of thermal and hydraulic interactions caused by fouling, flow and cleaning, and their otherwise hard to predict propagation through the network.
The significance of the work and applicability to other industries are briefly considered.
Biography
Sandro Macchietto, a graduate of Universita’ di Padova (Italy) and with Ms and PhD degrees from the University of Connecticut (USA), is Professor of Process Systems Engineering in the Department of Chemical Engineering, at Imperial College London, UK where he has provided research, project management and teaching leadership in chemical and process systems engineering for over 30 years. His research focuses on the development of methods for design, control and operations management of process and manufacturing systems and their application to sustainable energy systems. He pioneered methods for the modelling, simulation, optimisation and design of process systems and their application to a variety of industrial applications, from food to refineries, with an integrated, whole-systems approach. Some of his early optimisation methods have been widely used within 4 major simulation software. His work on model-based design of experiments for rapid model development has been highly cited in diverse application areas. He has co-authored over 150 papers and is a frequent keynote speaker at conferences.
At Imperial College Professor. Macchietto was Co-founder and then Director of the Centre for Process Systems Engineering (CPSE), co-founder and director of the Energy Futures Lab and launched and managed as Director a highly innovative MSc in Sustainable Energy Futures. He generated numerous collaborative projects at the interface between research and industry. He was Director and Chairman of Steering Committee of UNIHEAT, a UK-Russia multidisciplinary project addressing energy efficiency in refining (shortlisted for Research Project of the Year, IChemE Global Awards 2015). He co-founded and led two successful spinouts (Process Systems Enterprise Ltd – engineering software and services; Hexxcell Ltd – heat transfer software and services). He is on the Advisory Board of the ENSIACET (Institut National Politechnique Toulouse, France), the Scientific Committee of the Institut Mines Telecom in Paris, the Evaluation Board of the Politecnico di Torino (Italy) and various panels of the EU European Research Council (ERC).
Professional awards include the McRobert Award 2007, the top prize for innovation of the Royal Academy of Engineering, the 2009 IChemE Industry Awards for Innovation and Excellence, a Queen’s Award for Enterprise (with PSE), and a Queen’s Award for Excellence (with CPSE) and the 2016 Imperial College President’s Medal (the highest research award) for Research Excellence. Sandro was made a Cavaliere of the Order of Merit of the Italian Republic by the Italian President in 2003.