Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Strategic Energy Trading via Multi-Level Game Theory
The spread of renewable energy sources and energy storage created the need for novel interaction models that describe the exchange of energy and the acquisition of ancillary services in the energy grid. An energy market is a hierarchical structure in which the Distribution System Operator (DSO)designs the price of the energy and buys flexibility from the prosumers. The consumers selfishly aim at fulfilling their local demand while minimizing the cost. From this interaction stems the problem for the DSO of computing the the price and rewards that steer the prosumers’ behavior towards one that improves the efficiency of the grid. In this project, we aim at developing an equilibrium-seeking algorithm that relies on subgradient methods to compute the optimal polices for demand-side management on an electricity market with renewable energy sources.
Keywords: Hierarchichal Games, Equilibrium-Seeking Algorithms, Energy Markets
The European Commission highlighted in the Clean Energy Package that the presence of Renewable Energy Sources (RES) with variable generation patterns and the growing number of storage units create new market participants such as energy communities and aggregators. Due to this change, the electricity market is shifting towards smarter and more interconnected paradigms. In particular, the use of Distribution System Operators (DSO) is becoming a popular. The DSO works as an intermediate level between the national grid, i.e., Transmission System Operator (TSO), and the local grid composed of the flexible load, energy storage and RES. In this hierarchical structure, the DSO sells energy to the prosumers but, at the same time, it incentivizes them to provide ancillary services, e.g., flexibility, that are required by the TSO. Therefore, the DSO has to design the energy price used to sell the energy and the reward provided for the ancillary services.
This problem is often formalized in the literature as a Mathematical Programming with Equilibrium Constraints (MPEC). The policy selected by the DSO (the leader) is constrained by decision of the prosumers (the followers) that adapt their optimal energy demand based on the DSO's decision. The currently available solutions to solve MPECs usually relies on the mixed integer reformulation or iterative convexification}. Both these methods have several shortcomings, in fact they may be computationally expensive or require excessive information on the prosumers to be implemented. Furthermore, they has been developed to study bilevel optimization problems and cannot be easily extended to the more general case of a multilevel optimization problem.
Recent studies in the machine learning research community proved interesting convergence properties of subgradient methods applied to nonconvex and nonsmooth optimization problems. Motivated by these results, in this project, we aim at developing an subgradient-based algorithm that overcomes some of the limitations of the current solution methods for MPEC problems, and that can be extended also to multilevel optimization problems arising in modern energy systems.
The European Commission highlighted in the Clean Energy Package that the presence of Renewable Energy Sources (RES) with variable generation patterns and the growing number of storage units create new market participants such as energy communities and aggregators. Due to this change, the electricity market is shifting towards smarter and more interconnected paradigms. In particular, the use of Distribution System Operators (DSO) is becoming a popular. The DSO works as an intermediate level between the national grid, i.e., Transmission System Operator (TSO), and the local grid composed of the flexible load, energy storage and RES. In this hierarchical structure, the DSO sells energy to the prosumers but, at the same time, it incentivizes them to provide ancillary services, e.g., flexibility, that are required by the TSO. Therefore, the DSO has to design the energy price used to sell the energy and the reward provided for the ancillary services.
This problem is often formalized in the literature as a Mathematical Programming with Equilibrium Constraints (MPEC). The policy selected by the DSO (the leader) is constrained by decision of the prosumers (the followers) that adapt their optimal energy demand based on the DSO's decision. The currently available solutions to solve MPECs usually relies on the mixed integer reformulation or iterative convexification}. Both these methods have several shortcomings, in fact they may be computationally expensive or require excessive information on the prosumers to be implemented. Furthermore, they has been developed to study bilevel optimization problems and cannot be easily extended to the more general case of a multilevel optimization problem.
Recent studies in the machine learning research community proved interesting convergence properties of subgradient methods applied to nonconvex and nonsmooth optimization problems. Motivated by these results, in this project, we aim at developing an subgradient-based algorithm that overcomes some of the limitations of the current solution methods for MPEC problems, and that can be extended also to multilevel optimization problems arising in modern energy systems.
- Learn about Subgradient Methods and Bilevel Optimization;
- Formalize the problem of demand-side management in an energy market with TSO and DSO;
- Develop an subgradient-based equilibrium-seeking algorithm to compute the optimal policy;
- Validate your algorithm via numerical simulations and compare it with the state-of-the-art methods.
Publications: If the final results are promising they can potentially be turned into a publication.
- Learn about Subgradient Methods and Bilevel Optimization;
- Formalize the problem of demand-side management in an energy market with TSO and DSO;
- Develop an subgradient-based equilibrium-seeking algorithm to compute the optimal policy;
- Validate your algorithm via numerical simulations and compare it with the state-of-the-art methods.
Publications: If the final results are promising they can potentially be turned into a publication.
Please send your resume/CV (including lists of relevant publications/projects) and transcript of records in PDF format via email to gbelgioioso@ethz.ch and ccenedese@ethz.ch.
Start Date: Expected starting date is between January and March 2022, but it is negotiable.
Corona Disclaimer: This project can be done in person at the Automatic Control Laboratory (ifA), hybrid, or completely remotely. Most importantly, we can change between these forms as needed.
Please send your resume/CV (including lists of relevant publications/projects) and transcript of records in PDF format via email to gbelgioioso@ethz.ch and ccenedese@ethz.ch.
Start Date: Expected starting date is between January and March 2022, but it is negotiable.
Corona Disclaimer: This project can be done in person at the Automatic Control Laboratory (ifA), hybrid, or completely remotely. Most importantly, we can change between these forms as needed.