engleski

Towards a Steeper Learning Curve with the CrocodileAgent 2012

Since the liberalization and unbundling of power systems, power market simulators and agent-based simulation are gaining popularity. Power Trading Agent Competition (Power TAC) is a recent power market simulator that provides an opportunity to test market designs for smart grids. This paper describes the CrocodileAgent 2012 which is an attempt to build an efficient agent in the Power TAC competition. We demonstrate the interactions of modules within the agent and typical forecasting process that the CrocodileAgent 2012 performs for its needs. In the follow-up of those results, we describe two approaches that Agent 2012 used to exploit the drawbacks in the current market design: infinite periodic payment and PAC- MAN syndrome. Based on those, we propose more rounds in each competition cycle with an aim to increase the learning curve of Power TAC competitions and improve the speed of contribution to the society.

Smart grids ; Power TAC 2012 ; CrocodileAgent

nije evidentirano