MASTER THESIS Nr. 610 INTELLIGENT TRADING AGENT FOR POWER TRADING BASED ON THE REPAST TOOLKIT

1. MASTER THESIS Nr. 610INTELLIGENT TRADING AGENT FOR POWER TRADING BASED ON THE REPAST TOOLKIT Ivana Pranjić

2. Overview • Introduction • Agent based modeling • Electricity market • Smart grid • Agent-based modeling of electricity markets • Powertac • Repast toolkit • Implementing PowerTAC broker using Repast • Conclusion

3. Introduction • Most systems show signs of complex behavior • Complex adaptive systems – many interacting components • Adaptation and self-organization • Emergence • Non-linearity • Co-evolution • Simulation using agent-based models

4. Agent-based modeling • Suitable for imitating non-linear characteristics of complex adaptive systems • Agents – behaviors based on sets of rules • Autonomous • Able to learn and adapt • Interactive • Goal-oriented • ABM are flexible, provide natural description of complex systems and capture emergent phenomena

5. Electricity market • One of the most complex markets in the world • Centralized topology • transmission losses, power outages • Restructuring, liberalization • Distributed electricity generation - reliability, environmental and economical reasons • Introduction of the smart grid • Two-way flow of information and electricity • Integration of numerous new producers and encouraging use of renewable sources • Balancing power supply and demand in real-time

6. Smart grid

7. Agent-based modeling of electricity markets • Testing new market designs before implementing them in real life • ABM is commonly used for modeling electricity markets • Agents represent market participants who interact, make decisions, take actions and learn from their experience • There are numerous models of electricity markets • Power Trading Agent Competition (PowerTAC) – agents act as brokers and compete with each other

8. PowerTAC

9. Repast toolkit • Repast Simphony – a toolkit for agent-based modeling written in Java • Model agents using Java, Groovy, ReLogo or flowcharts • Graphical user interface – control the simulation, modify parameters, use external tools (Pajek, Weka) • Display data through charts and graphs • Libraries for neural networks, genetic algorithms, etc.

10. Implementing PowerTAC broker using Repast • “Sample broker” – agent template written in Java • Implementing the sample broker using Repast • Configure simulation parameters using the GUI • ClassNotFound and NoClassDefFound errors • Start the simulation without using the GUI • User is not able to control the simulation • Incompatibility between Repast development libraries and libraries used by sample broker

11. Implementing PowerTAC broker using Repast • Control the simulation execution manually • Give user the opportunity to manually create tariffs and wholesale offers and submit them to the market • PowerTAC simulation is managed by the simulator, not the agent

12. Conclusion • Agent-based models are often used to model electricity markets • Repast is not appropriate for implementing PowerTAC agent • Incompatible technologies • In Repast the user controls the simulation, in PowerTAC it is done by the simulator • Lack of documentation