A short-term load forecasting method is made based on the load curve shape similarity and human comfort index for distribution load characteristics. The characteristics of the load curve shape similarity are researched. Based on the expectation of each moment load difference, a load curve shape similarity calculation method is given. Based on the impact of meteorological factors on the load, a human comfort index is introduced, and according to the daily maximum temperature and minimum temperature, the similarity calculation method of a day human comfort index is given. According to the distribution load type characteristics, the load change characteristics of various types are researched on holidays and normal working days. Combining these features, this paper gives a distribution network short term load forecasting method, and discusses the method and its steps in detail. Nine days of historical data and weather data of a distribution network power point are taken as examples to verify the feasibility of the proposed method.