Con Edison is one of the largest energy companies in the United States, with approximately $13 billion in annual revenues (2016), and over $47 billion in assets. The company generates and delivers electricity to customers in New York City, Westchester County, and steam service to Manhattan.
Astor Wells worked on-site at their 4 Irving Place headquarters for Central Engineering on a number of projects. The following is a representative example.
Capital Asset Prioritization Tool with Machine Learning
Business Case
In the summer of 2006, the borough of Queens, New York experienced a series of blackouts. The cause was traced back to Con Ed’s decision to continue supplying power to 400,000 people after ten out of twenty-two feeder cables burnt out. An investigation determined that the failed feeder cables averaged 16 years, with the oldest 59 (see Wikipedia). The New York Public Service Commission severely criticized the company for its failure anticipate demand and/or equipment failure.
Con Edison engaged Columbia University’s Center for Computational Learning Systems (CCLS) to develop a decision support system that could be used to prioritize replacement of feeder cables. CCLS believed that a machine learning algorithm, armed with gigabytes of historical data, could be built to anticipate equipment failure.
Expertise
- Analysis & Design
- Database Design
- Software Development (VBA)
- Data Management & Lineage
Project highlights
The CAPT tool was a web-based front end to a series of programs running on a server with MS SQL Server as the database. Astor Wells was tasked to design a database for use by the machine learning algorithm. The model would “learn” by applying millions of what-if scenarios and arrive at an optimal solution. Each model “run” required recording of the input data set, the model version, and test results. Using database optimization techniques, AW was able to produce significant improvements in access times which gave the researchers the ability to test more models.