Deep Learning-Based Vulnerability Assessment of Distribution Grids Containing Distributed Power Sources and Electric Vehicles

A deep learning-based vulnerability assessment method for distribution networks containing distributed power sources and electric vehicles is proposed, which adopts a probability distribution to describe the uncertainty of distributed power sources, and describes the load model of electric vehicles with a travel demand model based on the constraints of the road network and a charging demand model considering the psychology of users. After considering the base load, a probabilistic tidal current model with source-load spatio-temporal characteristics is used for tidal current calculation. The LSTM neural network is used to predict the load voltage, and then the distribution network vulnerability assessment is carried out to obtain the node vulnerability model and the final output results. A branch circuit in Dandong is utilized for testing to verify the validity of this paper.