Modeling and Identification of Multi-Agent Systems with Applications to Smart Grids


Faculty Advisor: Prof Munther Dahleh and Dr. Mardavij Roozbehani
Contact e-mail:
Research Area(s): Control, Energy, Signals and Systems, Theoretical Computer Science
Understanding how smart appliances or electric vehicles respond to price signals is fundamental for stability and performance of future smart grids. For design and analysis purposes, it is important to develop low-complexity dynamical systems that model the aggregate response of a large number of agents (electricity consumers) to a common signal (e.g., price). This project builds on a previous Super UROP project that accomplished the following main tasks: 1. Implemented and simulated the dynamics of individual agents based on principles of dynamic programming. 2. Generated data by simulating the aggregate response of a large number of agents. 3. Applied system identification techniques to develop low order dynamical system models from the generated input/output data.
In this project we will be adding additional features and details to the model of individual agents and develop system identification and model reduction techniques for modeling the dynamics of the aggregate system. We will then use these models for design of efficient pricing mechanisms for matching supply and demand. The project requires basic skills and learning experience in control, dynamic optimization, system identification, and Matlab programming.

See more projects

Return to the SuperUROP site.