In collaboration with New York City's Power company, Con Edison, we aim to identify components of New York's underground electrical grid that are the most vulnerable to failure. This information can be used to assist with Con Edison's pre-emptive maintenance and repair programs. We focus in particular on prediction of manhole events (fires, explosions, smoking manholes) on the low-voltage network. These events can be difficult to predict, and some of the data used for the project are over a century old.
This project is the winner of the 2013 INFORMS Innovative Applications in Analytics Award.
21st-Century Data Miners Meet 19th-Century Electrical Cables. IEEE Computer.
Machine Learning for the New York City Power Grid. IEEE Transactions on Pattern Analysis and Machine Intelligence. (Spotlight Paper for the February 2012 Issue.)
A Process for Predicting Manhole Events in Manhattan. Machine Learning.