A learning approach for strategic consumers in smart electricity markets

In this paper we consider the design and the implementation of a machine learning approach and its integration with a widely used energy simulation platform. We focus on auction based energy markets which require their participants to bid for their energy demands or offers at small time intervals. Our agent based system utilize weather data to teach both consuming devices and renewable energy sources to bid in an effective manner. We simulate realistic case studies of a residential distribution power grid with a total of more than 600 households with varying energy requirements. Photovoltaic panels as well as wind turbines are the regional energy resources. Our experimentation exhibit the effectiveness of the learning procedure both in term of power consumption and cost.