Abstract: Electric vehicles (EVs) have the characteristics of large-scale deployment and high randomness, thus necessitating the establishment of their aggregated feasible region under uncertain conditions. This study proposes a scenario generation method for the aggregated feasible region of EVs based on multiple output Gaussian processes (MOGP). Firstly, training samples are generated directly from recent charging data to timely reflect the changing patterns of cluster charging behavior. Subsequently, to simultaneously capture the horizontal and vertical correlations of the feasible region envelope, a probability prediction method for aggregated feasible regions based on MOGP is proposed, along with an energy differential preprocessing method to maintain the monotonicity of the energy envelope of the EV cluster. Finally, the process of scenario generation and reduction is completed based on the trained MOGP model. Comparisons are made between the proposed method and existing methods using actual charging data for validation. Results indicate that the scenario set generated by the proposed method exhibits optimal accuracy and diversity and the generated energy envelope scenarios are more reasonable.