We analyze the impacts of future scenarios of electric vehicles (EVs) on the German power system, drawing on different assumptions on the charging mode. We find that the impact on the load duration curve strongly differs between charging modes. In a fully user-driven mode, charging largely occurs during daytime and in the evening, when power demand is already high. User-driven charging may thus have to be restricted because of generation adequacy concerns. In contrast, cost-driven charging is carried out during night-time and at times of high PV availability. Using a novel model formulation that allows for simulating intermediate charging modes, we show that even a slight relaxation of fully user-driven charging results in much smoother load profiles. Further, cost-driven EV charging strongly increases the utilization of hard coal and lignite plants in 2030, whereas additional power in the user-driven mode is predominantly generated from natural gas and hard coal. Specific CO2 emissions of EVs are substantially higher than those of the overall power system, and highest under cost-driven charging. Only in additional model runs, in which we link the introduction of EVs to a respective deployment of additional renewables, electric vehicles become largely CO2-neutral. (C) 2015 Elsevier Ltd. All rights reserved.