Line loss rate (LLR) is an important indicator of the service quality of power grid, and reducing line loss is the key to the sustainable development of electric power. Currently many researches focus on the LLR, however, most of them are based on theoretical data or a small amount of actual data, and rare studies on large scale actual data. In this paper, based on the data of China from 2013 to 2017, a comprehensive analysis in time domain and space domain of large-scale LLR is carried out through K-means clustering and temperature response model, including the time-domain evolution of LLR, the relationship between LLR and seasons, the composition of LLR level area, energy consumption characteristics and LLR distribution. It is found that, first, the annual LLR of China shows a downward trend, and the significant change cycle is similar to the per capita GDP. Second, the LLR in spring is significantly lower than that in other seasons, and shows a downward trend with the increase of temperature, which is consistent with the cumulative ownership ratio of air conditioning in that year. Third, in areas with high LLR, commercial users account for a large part of all users, and the proportion of users stealing electricity is significantly higher than that in general areas. Fourth, a novel healthy gradient evaluation index of LLR is established, and it is found that the distribution of LLR in unhealthy areas is close to normal distribution, while that in healthy areas is close to lognormal distribution. (C) 2021 Published by Elsevier Ltd.