The coal to electricity project changes the load characteristic of the power grid, and causes a significant impact on the line loss. In order to reduce the negative influence caused by coal-to-electricity project and improve the efficiency of the power supply unit, the low voltage district after the implementation of the coal-to-electricity project is studied, and then a method of line loss anomaly identification based on deep neural network is proposed.The proposed method establishes a model of line loss outlier identification combing the anomaly detection, EM algorithm and deep neural network. The model can predict whether the indicators of the low voltage district may cause the line loss be abnormal after implementing coal-to-electricity project, and then supply the guidance suggestions for the following coal-to-electricity project, so as to take corresponding measures to reduce the line loss.