With the increasing amount of data in smart grid, the existing power platform can not meet the demand of high performance computing. In order to solve this problem, cloud computing is introduced to design the memory computing framework of distributed power system, and intelligent meters and electronic devices are stored in distributed file system. combining the memory computing framework Spark and Newton-Raphson method, the Newton-Raphson method power flow calculation method based on elastic distribution data set (RDD) and directed acyclic graph (DAG) is proposed. Through the analysis of power flow calculation time, acceleration ratio and experimental cluster size, it is verified that the proposed algorithm has good performance. The research work can provide reference for the development of smart grid in China. This work is supported by National Key Scientific and Technological Project (No. 2016YFB0101900) and Science and Technology Project of Guangdong Power Grid Co., Ltd. (No. GDKJXM20161523).