Using MapReduce to Detect Anomalies in the Real-Time Smart Grid

Colin Hwang, Dylan Chamberlen, Michael Parros, John Russell, Jeremy Spruce, Aaron St. Leger, Suzanne J Matthews


Modern power grids quickly distribute electricity across large geographic areas with a high level of reliability. However, they are not invulnerable to widespread failures. In occasions of hardware failure, or fault (such as a transmission line tripping) the system can operate in a sub-optimal state and result in a loss of electric power to some customers. These events require grid operators to locate the point of failure in order to resolve the problem, a process which can take minutes or several days in large cascading blackouts. In recent years, engineers have explored ways to automate the rerouting process so that the grid can regulate itself. One such proposed system is the Real-Time Smart Grid, which seeks to monitor the vitals of a power grid in real-time. In this paper, we describe the anomaly detection software components of the Real-Time Smart Grid. Our solution incorporates the Phoenix++ MapReduce framework to process the large amount of data constantly produced by the grid in parallel. The algorithm enables the Real-Time Smart Grid to detect anomalies rapidly, provide data to automated controllers, and notify grid administrators of the location of any points of failure. This can enable grid operators to analyze and mitigate potential issues and concerns in a matter of seconds.


smart grid; anomaly detection; mapreduce; parallel computing

Full Text: PDF


  • There are currently no refbacks.