My friend Aaron Epel (Senior Data Scientist, Stem) and I talk about how machine learning is transforming fault prediction, asset failure prediction and preventive maintenance scheduling for energy grid operators.

00:00 Welcome and Intro

01:38 Meet Aaron, Senior Data Scientist at Stem

03:47 Aaron’s AI+ML origin story from IHS Energy

05:12 Masters in Operations Research, AI+ML

05:56 Predictive maintenance on wind turbines, Nexterra

07:05 Grid Operations Use Cases: Fault prediction, preventative maintenance scheduling

09:55 The huge costs of false positives and false negatives in grid operations

12:16 Why are AI+ML making such a big impact now in fault and asset failure prediction?

14:45 ML for complex faults that can only be captured with field data

15:57 Challenges for data science: very rare events, highly imbalanced datasets

18:41 Reframing fault and asset failure prediction as time-to-failure, survival analysis, RNN

23:18 The future of ML in fault and asset failure prediction

23:52 Inference at the Edge

25:22 Transfer learning from other domains

28:04 Wrap up and thanks Aaron!

Resources

James and Aaron electricity demand forecasting project: https://jamesaksanders.com/2023/10/20…

WTTE-RNN : Weibull Time To Event Recurrent Neural Network, EGIL MARTINSSON https://publications.lib.chalmers.se/…

WTTE-RNN – Less hacky churn prediction, EGIL MARTINSSON https://ragulpr.github.io/2016/12/22/…

Aaron Contact

Aaron Contact

https://www.linkedin.com/in/aaronepel/

aaron.epel@gmail.com