Limited by the power source mix and the grid structure restrictions, there are great challenges for the security operation of the grid and the consumption of the clean energy with the random and fluctuant wind power in the north, northeast and northwest parts of China (Three North regions for short) power system. Firstly, an analyzing method to divide and express different types of wind power fluctuation process is proposed by utilizing the self-organizing map (SOM) neural network algorithm in this paper. The research framework of wind power fluctuation characteristics considering multiple space and time scales is also proposed. Furthermore, on the basis of the actual operation data of the Three North regions in 2017, the short time characteristics and monthly/annual characteristics of the wind power fluctuations have been studied from three different spatial scales, i.e. the wind power plant, the provincial grid and the regional grid. From the analysis results, it can be seen that the wind power fluctuations categories can be efficiently identified by the self-organizing map clustering algorithm. The results also show that, the time and spatial scale fluctuation characteristic indexes of wind power can instruct the dispatching and operating decisions for the wind power enriched gird.