In this session, Con Edison will present on the project journey in developing a machine learning driven system powered by AMI data to identify open neutral conditions on the secondary distribution system. Open neutral conditions occurs when the neutral phase of the cable, is defective, oxidized or broken, which can result in service issues for the customer. However, more concerning is that open neutral conditions can be a source for electric shock which raises significant safety concerns. Con Edison is committed to providing a safe and reliable service for its customer and rectifying these safety issues are a top priority. The session will tell the story from idea inception to a fully deployed system which has to date successfully identified numerous open neutral conditions through the use of AMI voltage data and machine learning. While the importance of a robust data infrastructure and platform cannot be overlooked, a major ingredient to success was the close collaboration with the engineering team and the many months of iteration the project team undertook and will be a key theme throughout the session.

Session Takeaways:

• Overcoming the challenges of working with large unlabeled datasets
• Establishing early-on the process to ensure proper feedback loop for the machine learning model to continue to learn
• The importance of setting reasonable expectations, clear communication and the key needs to set-up a machine learning system