Abstract: Restricted by the natural conditions, the PV output is highly stochastic. In order to accurately assess the PV output characteristics of distributed PV power generation for rail transit infrastructure, a typical scenario generation method for distributed PV power generation for rail transit infrastructure based on the improved K-means clustering algorithm is proposed, and based on which the PV output characteristics are analysed. Firstly, based on the distributed PV power generation facilities as well as meteorological data, PV output data are simulated using PVsyst software. Then, for the problem of high blindness of the clustering parameters and initial clustering centre of the basic K-means clustering algorithm, it is improved by combining the density based index(DBI) index and hierarchical clustering, and a typical daily PV output scenario is generated by the improved K-means clustering algorithm. Finally, the effectiveness and superiority of the proposed method is verified based on the distributed PV system of rail transit infrastructure in central China, and the law and characteristics of the distributed PV output of rail transit infrastructure are revealed through qualitative and quantitative analysis of the output characteristics of each typical scenario.