In any conference the organizers face the problem of putting talks into sessions that should be preferably coherent in some way. And in many conferences there are talks that defy a proper classification to be put together. I was informed that this was the case with the talks in this session, which was a kind of residual category, for which the inventive name of Innovative Applications was come up with. However, in my view the talks in this session had several things in common (either application or techniques or even just `feel’ 😉 ) and surprisingly were quite coherent after all!
First Paul Bouman from the Erasmus University talked about the detection of activity patterns from smart card data. The application was transport demand. He talked about a method to deduce and analyse activity patterns and activity sequence patterns within the time dimension, based on smart card data. This method involves AI techniques such as clustering algorithms using a parametrized distance measure and a labeling procedure to label activities, leading to patterns which are visualized in some convenient way.
Next Markus Peters, also from Erasmus University, discussed the role of autonomous agents in future energy markets, and the 2012 Power Trading Agent Competition (Power TAC) in particular. This competition was set up to address limitations of earlier work of the use of agents in energy markets, viz. limitations in scope and limitations in competitiveness and comparability. The Power TAC platform includes simulator, broker agent framework, log analyzer and tournament manager, and is an open-source project, aimed at advanced students.
Finally Sara Ramezani (CWI, Amsterdam) also talked about AI solutions for energy markets, but at the level of how to protect LV cables in an energy grid from overheating, using smart battery operation. Apart from protecting the cables also optimizing the battery’s revenue is an important issue. Two robust heuristic online strategies, a purely reactive one and one that prepares for the future if possible by employing expectations of future prices, were proposed that did very well with respect to this optimization in simulations.
