A Bad Data Identification Method for Multiple Spatio-temporal Data in Power Distribution Network

Due to the excessive number of databases, unbalanced development and behindhand sensing infrastructures, distributed network data suffers from inconsistency, data missing, large measurement error and other data quality problems, which hinder the development of smart distribution network. In order to discover more complex deep-seated rules and provide more effective decision support for power system decision-making, it is necessary to study data mining and analysis methods that are suitable for massive data under current situation. This paper studies on the method of identifying bad data for multi-temporal and multi-spatial data in distribution networks and propose a method to identify bad data using likelihood-ratio test for 3D spatio-temporal data. In order to speed up the data processing rate, a 3D-LRT method based on multi-threading and Hadoop parallelization methods is proposed.