Leveraging MapReduce and Synchrophasors for Real-Time Anomaly Detection in the Smart Grid

Abstract

The rapid detection of anomalous behavior in SCADA systems such as the U.S. power grid is critical for system resiliency and operator response in cases of power fluctuations due to hazardous weather conditions or other events. Phasor measurement units are time synchronized devices that provide accurate synchrophasor measurements in power grids. The rapid deployment of PMUs enable improved real-time situational awareness to grid operators through wide area measurement systems. However, the quantity and rate of measurements obtained from PMUs is significantly higher than traditional devices, and continues to grow as more are deployed. Efficient algorithms for processing large-scale PMU data and notifying operators of anomalies is critical for real-time system monitoring. In this paper, we propose a novel, two-step anomaly detection approach that processes raw PMU data using the MapReduce paradigm. We implement our approach on a multicore system to process a dataset derived from real PMUs containing 4,500 PMUs ($\sim 18$∼18 million measurements). Our experimental results indicate the proposed approach detects constraint and temporal anomalies in under three seconds on 8 cores. Our work demonstrates the applicability of MapReduce for designing anomaly detection algorithms for the smart grid, and motivates the creation of novel MapReduce approaches for other SCADA applications.