With the integration of a large number of distributed energy sources, the operation and control methods of distribution systems have become increasingly complex. In response to the problems faced by distribution network state estimation methods such as difficulty in identifying distributed power source fluctuation data, low estimation accuracy, poor robustness and estimation timeliness, a distribution network distributed state estimation method based on integrated deep neural networks is proposed. First, the data identification technique of measuring data correlation testing is used to identify bad data and new energy fluctuation data. From this, the bad data is corrected using a temporal convolutional network (TCN) - bidirectional long short term memory (BILSTM). Then, an integrated deep neural network (DNN) state estimation model is established, and the maximum relevance-minimum redundancy (MRMR) method is used to optimize the training samples, thereby improving accuracy and robustness. Finally, a distributed integrated DNN model is established to compensate for the slow speed of centralized state estimation and improve efficiency. The numerical analysis based on an IEEE123 distribution network shows that the proposed method can more accurately identify distributed power source fluctuation data and bad data, while improving the accuracy and efficiency of state estimation, and is very robust.