Abstract: Aiming at the problem of slow solving speed of existing distribution network（DN）reconstruction algorithm,a fast distribution network reconfiguration method based on convolutional neural network（CNN）is proposed.Firstly,a multi-branch CNN model based on the loop structure of DN is established,which can reduce the dependence on the concrete structure of DN in the process of modeling.Secondly,the CNN model is trained based on the hybrid training method,so that the model has the ability to quickly reconfigure the DN with different load modes. Finally,the performance of the model is analyzed in IEEE 33-bus system to verify the effectiveness of the proposed method.