Abstract: The proposal and development of smart grid corresponding brought a large number of inputs of power quality monitoring terminals in the regional power grid. Meanwhile, the increase of sampling frequency and the monitoring time has accelerated the explosion growth of sampling data in the power quality monitoring platform. In the traditional power quality monitoring platform, the monitoring terminal data uploaded and processed by adopting centralized mode. With the advancement of the smart grid construction, higher requirements of computing speed and precision of the power quality monitoring indexes are put forward. The storage capacity and computing ability of the server in traditional power quality monitoring platform can hardly meet the growing demands of operation requirements.Though purchasing a higher configuration server can temporarily meet the demand of computing, it will lead to large waste of resources when there is no running tasks. The purpose of this study is to realize a high efficiency calculation of sampling data of electric power system under low hardware cost and minor resource waste. It is possible to make full use of the powerful data storage capacity and computing ability under Hadoop distributed file system and parallel programming model to calculate the basic sampling data in power quality monitoring platform; Based on the study of windowed interpolation theory of Fourier analysis method, a novel kind of improved sidelobe characteristics window function -- time domain multiplication window can enhance the accuracy of calculation. The harmonic analysis algorithm based on time domain multiplication window is applied to the MapReduce framework of the power quality monitoring platform. The feasibility and superiority of the proposed parallel processing model of the power quality platform are verified through the experiment based on a small Hadoop cluster.