By using statistical analysis technique

the difficulty of lacking causality in short-term wind speed prediction can be overcome by extrapolating the known time series. However

because of the fuzziness of subjective cognition

challenges exist in the choice of extrapolation models

parameters and training samples. To reduce the influences of fuzziness of subjective cognition on the performance of classification prediction and improve the efficiency of sample classification

the concept of coarseness of wind speed time series(WSTS)is proposed. Symbols defined according to tendency features are used to describe WSTS. Based on this

a two-layer symbolizing method using unit window feature and variation trend feature is proposed to improve WSTS symbolization. Finally

a case study based one year data collected from a wind farm at Jiuquan wind power base in Gansu Province is presented to validate the effectiveness of the proposed coarseness method.