Abstract: Given the complexity and growing collection scale of the distribution system big data,it is urgent to figure out how to effectively preprocess data in distribution system. Considering that,this paper proposes a parallel computing technology for large-scale datasets based on Apache Spark. Firstly,the Apache Spark-based big data parallel computing platform is set up to improve preprocessing efficiently. Then,we analyze some common problems in the present distribution system big data and put forward data governance programmes accordingly. After that,we introduce the specific processes of distribution network big data preprocessing combined with the Spark computing engine. Finally,the experiments verify that the proposed data preprocessing method improves the distribution system in data prediction accuracy and that the distributed computing platform possesses high speed in data preprocessing.