We are pleased to have Dr. Matthijs Spaan as the invited speaker at MSDM-2015.
Computational Challenges for Scheduling in Active Distribution Networks
Abstract: Realizing a high penetration of renewable energy technologies in distribution grids is accompanied by important challenges. The integration of distributed generation, such as solar panels, may produce both voltage dips and local voltage peaks. Additionally, the low marginal cost of renewable electricity may lead to congestion if low prices coincide with peak demand, especially when large loads respond en-masse to real-time price signals. Traditionally, such problems have been tackled through grid reinforcements, which is a very costly approach. A promising alternative is to actively manage power flows by involving the local users, thereby transforming the grid into an active distribution network. We identify the main computational challenges in the management of such active distribution networks, with a focus on local congestion management and voltage control. Firstly, when optimizing power flows in distribution networks, new computational challenges arise on top of those in transmission networks, as simplifying assumptions on power losses, voltage levels, and balanced phases no longer hold. Secondly, new challenges come from the scheduling of flexible loads, because of heterogeneous characteristics, uncertainty, communication problems, and scalability. Thirdly, involving users in distribution network management introduces additional communication and computational challenges, enlarged by the physical interdependencies and uncertainty in distribution networks.
Dr. Matthijs Spaan is an assistant professor at the Algorithmics group, Delft University of Technology, Delft, The Netherlands. Prior, he was a senior research scientist at the Intelligent Robot and Systems Group, Institute for Systems and Robotics, Instituto Superior Técnico, Lisbon, Portugal. His research interest is in planning under uncertainty for multiagent and multi-robot systems, in particular, the development of theory and solution methods for decentralized partially observable Markov decision process (Dec-POMDPs). He has applied approximate POMDP planning techniques to robotic applications, such as Network Robot Systems. His group demonstrated successful POMDP-based cooperation between surveillance cameras and mobile robots. Dr. Spaan received his PhD degree in Computer Science (2006) and an MSc degree in Artificial Intelligence (2002), both from the University of Amsterdam, The Netherlands.