Data Science Community
In its simplest form, data science involves developing methods of recording, storing, and analyzing... View more
Organizer:
- Organized by
-
-
Upcoming May (2022) Community Conversation – In-Person at UA Summit!
-
Upcoming May (2022) Community Conversation – In-Person at UA Summit!
Hello Data Science Community Members,
I hope all is well! We are just under one week away from our In-Person May (2022) Data Science Community Conversation taking place live at UA Summit 2022. On Tuesday, May 10th at 1:30 PM CT, we’ll be joined by Brad Gall, Sr Data Architect from The Energy Authority who will be leading a discussion around, “Using Machine Learning to Detect Broken AMI Meters.”
Please note this session will only be taking place live at UA Summit, no virtual option will be available. If you have not yet registered for UA Summit there’s still time! Please use the link below and reach out to me at kpraet@utilityanalytics.com if you have any questions or need help registering.
https://www.utilityanalyticssummit.com/2022/1486114
Learn more about this session and our speaker below!
Session Description:
We all know AMI systems are great at collecting detailed metering information remotely. But without regularly scheduled field checks, how do you determine if your meter reads are still accurate over time? Learn how TEA’s Connected Analytics group is using machine learning to find broken water meters, reduce unnecessary field activity, and ultimately protect the utility’s revenue.
Speaker: Brad Gall, Sr Data Architect(The Energy Authority)
Brad Gall is a Cloud Data Architect with over 20 years’ experience in Information Technology and Data Analytics.Working primarily with Microsoft data platforms, Brad has successfully delivered data driven projects across multiple industries in both corporate and consulting roles.
_____________________________________________________________________________________________________________________________________________________
This session will again only be taking place live at UA Summit. If you have any questions prior or need help registering, please let me know.Thanks!
——————————
Kevin Praet
Membership Coordinator
Utility Analytics Institute (UAI)
Boulder CO
315-440-3033
—————————— -
2 Replies
-T.jpeg)
-
SDG&E took a similar approach but we didn’t have enough interval and voltage data to productionalize. Once we upgrade the smart meter infrastructure to the next generation, this type of analysis will be possible.
——————————
Rob Malowney
Asset Data Systems & Records Manager
SDG&E
san diego CA
6196196196
——————————
——————————————-
Original Message:
Sent: 05-03-2022 10:52
From: Tom Mahar
Subject: Transformer Failure via AMI Voltage DataHi Jordan,
I know ComEd had done it using 30 minute AMI voltage data, plus some additional features which described the transformer (assuming you’re talking about service transformers). Per some of our technical experts, sometimes when a transformer begins to fail it causes a step change of voltage at the AMI meters which is why voltage is a key input.
However we found that this project offers weak benefits. For example, let’s say your algorithm turns out OK with a precision of 33%. When you crunch the numbers the cost of replacing 3 transformers in order to prevent a single small outage just might not work out in most cases. If there’s an existing funded program for replacing transformers this math might be different, but trying to justify pulling funds away from more impactful reliability programs in order to fund something like this is a difficult task.
——————————
Tom Mahar
ComEd Advanced Analytics Team
——————————Original Message:
Sent: 05-02-2022 13:34
From: Jordan Pino
Subject: Transformer Failure via AMI Voltage DataHi UAI,
We are working on a predictive model using AMI interval voltage data to predict transformer failures. However, we are struggling majorly I suspect due to inadequate data availability (~18 months).
Just curious as to if anyone else has done this (predicted transformer failures using AMI data)…..how much data did you use to train your model? How did it turn out?
Thanks,
Jordan
——————————
Jordan Pino
Manager – Data Science
Entergy
The Woodlands TX
—————————— -
I agree with Tom’s points…hence why many utilities essentially ‘run them to failure’.
——————————
Eddie Fee
Director, Metering & Field Ops
Orlando Utilities Commission
Orlando FL
407.434.2262
——————————
——————————————-
Original Message:
Sent: 05-03-2022 10:52
From: Tom Mahar
Subject: Transformer Failure via AMI Voltage DataHi Jordan,
I know ComEd had done it using 30 minute AMI voltage data, plus some additional features which described the transformer (assuming you’re talking about service transformers). Per some of our technical experts, sometimes when a transformer begins to fail it causes a step change of voltage at the AMI meters which is why voltage is a key input.
However we found that this project offers weak benefits. For example, let’s say your algorithm turns out OK with a precision of 33%. When you crunch the numbers the cost of replacing 3 transformers in order to prevent a single small outage just might not work out in most cases. If there’s an existing funded program for replacing transformers this math might be different, but trying to justify pulling funds away from more impactful reliability programs in order to fund something like this is a difficult task.
——————————
Tom Mahar
ComEd Advanced Analytics Team
——————————Original Message:
Sent: 05-02-2022 13:34
From: Jordan Pino
Subject: Transformer Failure via AMI Voltage DataHi UAI,
We are working on a predictive model using AMI interval voltage data to predict transformer failures. However, we are struggling majorly I suspect due to inadequate data availability (~18 months).
Just curious as to if anyone else has done this (predicted transformer failures using AMI data)…..how much data did you use to train your model? How did it turn out?
Thanks,
Jordan
——————————
Jordan Pino
Manager – Data Science
Entergy
The Woodlands TX
——————————
Log in to reply.