Prosumer clustering into virtual microgrids for cost reduction in renewable energy trading markets

Today's electricity system is undergoing a transformation from a model of centralized electricity generation, to a more decentralized paradigm, where a large number of small energy prosumers (i.e. both producers and consumers) generate energy and may participate in the energy market. In these markets, energy is usually traded at a time prior to the time of delivery in an exchange, based on forecasts of the production and the consumption, and the cost for the prosumers is related to the accuracy of these forecasts, through the application of penalties when an imbalance appears. In this paper we study the problem of orchestrating the energy prosumers into virtual clusters, in order to participate in the market as a single entity and to reduce the total energy cost, through the reduction of the total relative forecasting inaccuracies. Using a real dataset of 33 prosumers located in Greece, we study different clustering algorithms, including spectral, genetic and an adaptive algorithm. The performance evaluation results show that significant cost reduction may be achieved, through the association of the prosumers into groups.