Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Decomposition methods for large-scale energy systems under uncertainty
In this project, we aim to derive decomposition methods for large-scale energy systems optimization problems under uncertainty.
Keywords: Optimization under uncertainty, two-stage stochastic programming, energy hubs
Energy hubs are multi-generation systems
where multiple energy carriers are converted,
stored and supplied. With an increasing share
of renewable energy sources, largely decentralized
and of intermittent nature, energy hubs are
gaining relevance in the energy landscape as
promising solutions because they match local
production with consumption. To support their
widespread adoption, a techno-economic analysis
is necessary: “what technologies should be
installed?”, “how to size them to match the local
demand?”, “where to optimally locate them?”
are only few of the key questions behind energy
hubs design task. The optimal design of energy hubs is a difficult task due to 1) the presence of
uncertainty (such as, weather conditions or local energy demand) affecting the system and 2) its
large-scale nature. The problem is typically cast as a “two-stage” program, where the “here-andnow”
first stage decisions concern the design phase, while the “wait-and-see” second stage ones
concern the operations phase. The main goal is to minimize the total cost, while maximizing its
robustness with respect to uncertainty in the problem parameters.
Recently, the paradigm of distributionally robust optimization (DRO) has gained momentum
owning it to the ability to robustify against uncertainty in the distribution of the stochastic parameter
by considering the worst-case distribution in a so-called ambiguity set. Specifically, we focus on
data-driven DRO under the Wasserstein distance and aim to solve a “two-stage distributionally
robust optimization problem” assuming the availability of historical realizations of the uncertainty.
In this project, we aim to explore and compare different methods to reduce the computational
footprint of the DRO program in two directions. On one side, we resort on moving-block strategies
to reduce the overall number of optimization variables. On the other side, we wish to cluster the
available disturbance data to reduce the number of constraints of the DRO problem. Our final goal
is to establish a reduced-order framework to solve the energy hub design problem with statistical
guarantees and theoretical a-priori certificates on the incurred suboptimality gap.
Energy hubs are multi-generation systems where multiple energy carriers are converted, stored and supplied. With an increasing share of renewable energy sources, largely decentralized and of intermittent nature, energy hubs are gaining relevance in the energy landscape as promising solutions because they match local production with consumption. To support their widespread adoption, a techno-economic analysis is necessary: “what technologies should be installed?”, “how to size them to match the local demand?”, “where to optimally locate them?” are only few of the key questions behind energy hubs design task. The optimal design of energy hubs is a difficult task due to 1) the presence of uncertainty (such as, weather conditions or local energy demand) affecting the system and 2) its large-scale nature. The problem is typically cast as a “two-stage” program, where the “here-andnow” first stage decisions concern the design phase, while the “wait-and-see” second stage ones concern the operations phase. The main goal is to minimize the total cost, while maximizing its robustness with respect to uncertainty in the problem parameters. Recently, the paradigm of distributionally robust optimization (DRO) has gained momentum owning it to the ability to robustify against uncertainty in the distribution of the stochastic parameter by considering the worst-case distribution in a so-called ambiguity set. Specifically, we focus on data-driven DRO under the Wasserstein distance and aim to solve a “two-stage distributionally robust optimization problem” assuming the availability of historical realizations of the uncertainty. In this project, we aim to explore and compare different methods to reduce the computational footprint of the DRO program in two directions. On one side, we resort on moving-block strategies to reduce the overall number of optimization variables. On the other side, we wish to cluster the available disturbance data to reduce the number of constraints of the DRO problem. Our final goal is to establish a reduced-order framework to solve the energy hub design problem with statistical guarantees and theoretical a-priori certificates on the incurred suboptimality gap.
The goals of the project are as follows:
1. Learn about Two-stage Stochastic Optimization, Distributionally Robust Optimization, Energy
hubs;
2. Develop decomposition schemes for large-scale DRO problems with statistical guarantees;
3. Validate the designed algorithm via numerical simulations on the problem of energy hub design.
The goals of the project are as follows:
1. Learn about Two-stage Stochastic Optimization, Distributionally Robust Optimization, Energy hubs;
2. Develop decomposition schemes for large-scale DRO problems with statistical guarantees;
3. Validate the designed algorithm via numerical simulations on the problem of energy hub design.
If you are an highly motivated student with interest in optimization under uncertainty, Please send your resume/CV (including lists of relevant publications/projects) and transcript of
records in PDF format via email to mfochesato@ethz.ch.
If you are an highly motivated student with interest in optimization under uncertainty, Please send your resume/CV (including lists of relevant publications/projects) and transcript of records in PDF format via email to mfochesato@ethz.ch.