The active distribution network operation situation prediction is an important tool to guarantee the safety and stability of the distribution network and the hazard perception. For the fast and accurate prediction of active distribution network operation, this paper proposes an active distribution network short?term operation prediction method based on ICEEMDAN?TA?LSTM model. Firstly, the original sequence is decomposed into several stable time series components by improving the modal decomposition to reduce the irregularity of the original data. At the same time, the improved manta ray feeding optimization algorithm is used to optimize the model's hyperparameters to comprehensively improve the overall prediction accuracy of the model. Then, from the perspective of nodes and branches, the node voltage overrun margin, branch load severity and voltage/current fluctuation evaluation indexes are proposed to characterize the distribution network operation situation at multiple levels. Finally, the feasibility and effectiveness of the model proposed in this paper are verified by taking the improved IEEE 33 node as a typical calculation example.