Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Game-theoretic MPC for Smart Homes
We aim to design a novel control approach which combines principles of Game Theory and Model Predictive Control. As part of a collaboration with tiko AG, we then want to apply the new controller for coordination of a large number of smart homes with power suppliers.
Keywords: model predictive control, system, identification, game theory, power systems, demand side management
An ever an increasing number of engineering applications (large-scale infrastructure, autonomous driving, drone flights) require real-time coordination of multiple control algorithms which selfishly optimise for their own benefit. We develop game-theoretic control algorithms to take into account interactions between so-called self-interested agents. Specifically, in the spirit of model predictive control (MPC), we want optimally control and coordinate actions of multiple taking into account future predictions and constraints.
Our current power grid is under increasing stress caused by climate change and the volatility of renewables which already results in a growing number of blackouts (in the US) during peak demand periods. One approach to decrease peak demand is to incentivise users to reduce consumption in emergency situations through demand side management (DSM). In this project we collaborate with tiko AG to test our game-theoretic control in a real world application.by implementing DSM. tiko AG is a company that retrofits household devices (boilers, heaters, photovoltaic devices) to create smart homes. By coordinating a large number of these smart homes and applying DSM, it is possible to reduce peak power consumption in a region over a certain time period.
However while shifting the load, it is important to prioritise the interests of prosumers, e.g temperature comfort ranges and energy cost, otherwise they might opt-out of the program. Thus our objective is to implement a game-theoretic control approach which takes care of the individual interests of all prosumers while incorporating future predictions and constraints of the grid.
**Highlights of this project:**
- Understand and work with novel control approach relevant for many real-world applications
- Perform system identification on real data collected in over thousands of households
- Work on solving a highly relevant problem of managing our future power consumption in a smart way
- The proportions of practical and theoretical work can be adapted along the way depending on the student's interests and the course of the project
An ever an increasing number of engineering applications (large-scale infrastructure, autonomous driving, drone flights) require real-time coordination of multiple control algorithms which selfishly optimise for their own benefit. We develop game-theoretic control algorithms to take into account interactions between so-called self-interested agents. Specifically, in the spirit of model predictive control (MPC), we want optimally control and coordinate actions of multiple taking into account future predictions and constraints.
Our current power grid is under increasing stress caused by climate change and the volatility of renewables which already results in a growing number of blackouts (in the US) during peak demand periods. One approach to decrease peak demand is to incentivise users to reduce consumption in emergency situations through demand side management (DSM). In this project we collaborate with tiko AG to test our game-theoretic control in a real world application.by implementing DSM. tiko AG is a company that retrofits household devices (boilers, heaters, photovoltaic devices) to create smart homes. By coordinating a large number of these smart homes and applying DSM, it is possible to reduce peak power consumption in a region over a certain time period.
However while shifting the load, it is important to prioritise the interests of prosumers, e.g temperature comfort ranges and energy cost, otherwise they might opt-out of the program. Thus our objective is to implement a game-theoretic control approach which takes care of the individual interests of all prosumers while incorporating future predictions and constraints of the grid.
**Highlights of this project:**
- Understand and work with novel control approach relevant for many real-world applications
- Perform system identification on real data collected in over thousands of households
- Work on solving a highly relevant problem of managing our future power consumption in a smart way
- The proportions of practical and theoretical work can be adapted along the way depending on the student's interests and the course of the project
1. Learn about game theoretic control (i.e. feedback equilibrium seeking, receding horizon games, game theoretic planning) and the DSM application
2. Perform system identification on data provided by tiko AG to construct a model of a distribution grid and thermal models of smart homes
3. Mathematically formulate the game-theoretic control problem and solve it with an appropriate algorithm
4. Test your strategy in closed-loop with your model
1. Learn about game theoretic control (i.e. feedback equilibrium seeking, receding horizon games, game theoretic planning) and the DSM application
2. Perform system identification on data provided by tiko AG to construct a model of a distribution grid and thermal models of smart homes
3. Mathematically formulate the game-theoretic control problem and solve it with an appropriate algorithm
4. Test your strategy in closed-loop with your model
Please send your resume/CV (including lists of relevant publications/projects) and transcript of courses and grades in PDF format via email to gbelgioioso@ethz.ch, shall@ethz.ch, dliaomc@ethz.ch and mbadyn@ethz.ch.
Please send your resume/CV (including lists of relevant publications/projects) and transcript of courses and grades in PDF format via email to gbelgioioso@ethz.ch, shall@ethz.ch, dliaomc@ethz.ch and mbadyn@ethz.ch.