Electric vehicles embedded virtual power plants dispatch mechanism design considering charging efficiencies

The increasingly popular electric vehicles (EVs) are changing the control paradigm of the power grid due to their uncoordinated charging behaviors. However, if well coordinated, smart homes, workplaces, and other locations that support EV charging could provide the grid with the urgently required flexibility via virtual power plants (VPP). In this paper, we develop the EV charging schedule model by capturing the unwillingness of EV drivers to alter their initial charging behaviors, referred to as the discomfort function. Predictability and the value of charging time, which represent the electricity consumption stability and the time value of EV drivers, characterize the discomfort function. Rather than existing works capturing discomfort by a direct simple parameter, such a computable data-driven quantification of discomfort enables us to customize an efficient VPP dispatch mechanism for EVs. In addition, to deal with the unknown charging efficiencies of EVs, we apply chance constraints only with the knowledge about mean and standard deviation of charging efficiencies, rather than their specific distribution. Using the concept of conditional value-at-risk (CVaR), we build an effective algorithm to solve the practical non-convex VPP dispatch model considering charging efficiencies. The effectiveness of our proposed models and associated algorithms are validated by numerical studies.