Autonomous agent-based approaches to the Broker Problem in Electricity Markets



The coordination between customers and producers of electricity is becoming increasingly decentralized. Vertically-integrated utilities traditionally ensured that enough electricity is generated by large, central power plants to exactly meet customers' electricity demand in real-time. But liberalization efforts and the advent of decentralized, small-scale and/or intermittent power sources are challenging the traditional solution to the coordination problem.
The hope is for market mechanisms to fill in. Brokers in these markets are facing the challenging task of intermediating between customers with uncertain demand and large-scale producers while ensuring approximate real-time balance between the two. Limited visibility of customer preferences, strategic behavior of competitors, and exogenous influences such as the weather further exacerbate the brokers' problem.
Can autonomous learning agents be successful in such an environment?
In this discussion we will revisit the broker problem, highlight possible starting points for research into broker agents, and explain why the combination of reinforcement learning with methods from preference modeling appears to be one attractive solution strategy.
Contact information:
Wolf Ketter