In order to determine the short-term change trend of distribution line loss rate and meet the need of line loss lean management, this paper proposes a distribution daily line loss rate forecasting model based on a combination of Denoising Auto-Eencoder (DAE) and Long Short-Term Memory (LSTM) network. First, this paper establishes a grey comprehensive correlation indicator to analyze the correlation between the recent sequences of daily line loss rate influencing factors and the same period in the previous year. The influencing factors of the previous year are added to the model input to assist prediction. Then, to realize the feature extraction and dimension reduction of input sequences, the DAE is constructed to encode and reconstruct the input in an unsupervised way. Finally, the coded sequences are input into the LSTM and the model is obtained by training. A case analysis is carried out using the data of multiple distribution lines in a city of Hunan. The results show that the proposed model has superior accuracy and moderate calculating speed. This has a certain engineering application value. This work is supported by the International Science and Technology Cooperation Program of China (No. 2018YFE0125300), the National Nature Science Foundation of China (No. 52061130217), the Innovative Construction Program of Hunan Province of China (No. 2019RS1016) and the Excellent Innovation Youth Program of Changsha of China (No. KQ1905008).