Welcome to the research group of Alexander Mitsos
Our research focus is the optimization of energy systems, including the projects described below.
Our research relies on modeling, simulation and optimization.
We use commercial as well as in-house software packages, both sequential and parallel.
There are several opportunities to join the group.
Microchemical systems have attracted academic and industrial interest because they allow reactions to be performed under aggressive conditions and because often higher yields than in conventional reactors can be achieved.
Another area with a potential for microchemical systems is that of consumer products, such as lab-on-chip applications or man-portable power generation.
In the latter case, the main benefit of microchemical systems is the miniaturization of the end product, as opposed to a performance increase.
Our research considers the optimal system design.
Ongoing work is improving this methodology and generalizing it to other micro-chemical systems.
Duality in thermodynamics
Classically, the stability of phase equilibria is established by the well-known Gibbs tangent plane criterion, i.e., that a phase split is stable if and only if all phases lie on the same tangent to the Gibbs free surface and the tangent plane is below the Gibbs free energy surface for all compositions.
Recently, we proposed an alternative interpretation as a dual stability criterion.
Using the geometric interpretation of Lagrangian duality and its similarity to the Gibbs tangent plane criterion, we showed that nature goes to the solution of the Lagrangian dual problem and (in the presence of duality gap) not to the solution of the primal problem, as one would expect.
Ongoing research is to generalize this exciting theoretical contribution to other equilibrium systems.
Parameter Estimation in Phase Equilibria
The accurate prediction of equilibria is of tremendous importance in chemical and energy systems.
Models such as NRTL or UNIQUAC contain adjustable parameters that describe the nonideality of mixtures.
Provided that suitable parameter values are used, the models can accurately predict the chemical potential of most liquid mixtures.
Parameter estimation methods have been around for many decades, however conventional methods employ necessary only stability criteria, and as a consequence qualitative and quantitative errors in the prediction can occur.
Recently, we proposed a formulation for liquid-liquid equilibria and vapor-liquid-liquid equilibria (in press with Chemical Engineering Science) based on bilevel programs method.
Ongoing work is to generalize it to more complicated cases (in collaboration with George M. Bollas)
and to provide a robust implementation for the use by the academic community.
Solar Thermal Energy
A major challenge for man-kind in the coming years will be to meet the growing energy demands at the same time that natural resources become scarcer and carbon dioxide emissions need to be put under control.
Renewables and particular solar energy need to be investigated.
We are exploring the optimization of a novel solar-thermal process in collaboration with
Alexander H. Slocum (MIT),
John G. Georgiadis (UIUC),
and the Energy, Environment and Water Research Center (EEWRC) at The Cyprus Institute.
Authorities in Cyprus also participate in this project.
This project has already received attention from the local press, e.g.,
Alithia,
Simerini and
Politis.
Reaction Networks and Reactive Flows
Simulating reactive flows, e.g., for design of internal combustion engine,
is very challenging because these systems combine transport phenomena and chemistry.
Our interests include model reduction and combined optimization of engines and exhaust-gas after-treatment devices.
Optimization Algorithms
Numerical optimization (Mathematical Programming) is a tremendously useful tool for engineers.
We extensively use optimization algorithms for solving the above described engineering problems.
Robust and reliable algorithms and commercial software packages exist for many formulations.
However, significant algorithmic development is still required in other cases, in particular embedded programs (bilevel and SIP) under nonconvexity.
Moreover, the use of parallel architectures has not been investigated sufficiently, in particular in global optimization.