Brought to you by NumFOCUS Foundation and O’Reilly Media
The official Jupyter Conference
Aug 21-22, 2018: Training
Aug 22-24, 2018: Tutorials & Conference
New York, NY

Anomaly detection and classification with distribution grid sensor data

Who is this presentation for?

  • Data scientists and researchers

Prerequisite knowledge

  • A basic understanding of probability, Python, and classification algorithms

What you'll learn

  • Understand the problems facing the distribution grid and the current technologies being used to address them
  • Learn how probability theory can aid in anomaly detection and how to assess the accuracy of classification algorithms


Alaa Moussawi offers an overview of anomaly detection algorithms that use data from phasor measurement units (PMUs) that have been recently installed throughout the distribution grid in an effort to improve efficiency in monitoring and maintenance of the grid.

These algorithms are probabilistic in nature and assume a basic set of dynamics guiding the system. The implementation is heavily reliant on basic statistical functions within the Jupyter environment. Anomalies are classified using fundamental classification algorithms such as decision trees and neural networks. Feature selection is used to identify the optimal set of parameters to be ingested by the learning algorithms, and the accuracy of the algorithms is assessed to determine which pipeline results in the most efficient use of the data for the assessment of the state of the grid, utilizing learning algorithms and supplementary functions from the scikit-learn and Keras libraries.