Managing Energy in a Virtual Power Plant Using Learning Classifier Systems

Many renewable energy resources like windturbines or solar collectors are volatile and unstable. Distributed energy producing entities are clustered to virtual power plants that – from a black-box perspective – should act like conventional stable power plants. It is the task of intelligent control strategies to compensate fluctuations and to exploit renewable recourses to a maximal extend. In this paper we present a simple simulation model of a virtual power plant that is based on real-world power consumption and wind production data. A power storage system allows to save overcapacities and to supply energy in case of demand bursts. A reserve power plant allows additional energy production to compensate shortcomings. The simulated power plant can easily be extended by parameterizable modules. We use a learning classifier system to evolve rules that control the energy storage system and the reserve plant to compensate energy fluctuations. A study of evolutionary parameters as well as a discussion of the evolved rules complement the experimental analysis. A simple demand side management example demonstrates the influence of pricing on the virtual power plant.