In view of the problem that the time series similarity model based on Euclidean distance is less efficient and incomplete in morphology, the paper presents a similar model based on trend for time series. The model is based on piecewise linear representation and symbolic representation of time series, and the difference of time series is obtained by calculating the edit distance of the symbol string. At the same time, the model considers the time series length and the time series variation separately, expanding the use of the model. The experimental results show that the similarity model of time series based on trend is suitable for the similarity matching, and the efficiency is higher than traditional model which is based on Euclidean distance. Since the model emphasized on the similarity of morphology, it’s not sensitive to the deviation of time axis and white noise, and it’s suitable for the similarity matching of time series with different length.