Based on the kernel density function

a charging load calculation model with coupling characteristics is improved. In this model

the kernel density function instead of deterministic probability distribution function is selected to fit electric vehicles’ (EVs’) driving characteristics

the multi-dimensional probability distribution function to generate EVs’ driving random numbers with coupling features

and the charging probability to describe users’ uncertainty charging behavior. By comparing the calculated charging load with the actual charging load of a normal EVs’ charging station

the superiority of this method is verified. And then based on the improved model

the charging load of future large-scale residents’ private EVs is predicted. It is shown by research that the uncertain charging behavior exerts a great impact on EVs’ charging load. With decreased charging probability

the peak of charging load is greatly reduced

the peak value variation range and charging duration both increased and uncertain effects are brought to the grid load. This work is supported by National High Technology Research and Development Program of China (863 Program) (No. 2011AA05A109).