Steve Mitnick is President of Lines Up, Inc., Editor-in-Chief of Public Utilities Fortnightly, and author of “Lines Down: How We Pay, Use, Value Grid Electricity Amid the Storm.”
PUF: What are we looking at here?
Paul Hofmann: This demo is about improving capacity utilization and generation for utilities. We use a wind farm to show how to improve the installed capacity. We start with a dashboard. Using it we can do a deep dive to see the maximum capacity that we could achieve.
We have a natural loss of capacity because the wind doesn't blow all the time. Then we have a technical loss because the windmill is not up all the time — it has to be repaired or updated.
Now, let's drill down into the dashboard and see what's going on. Here we have a predictive maintenance tool, where we can look into the future and estimate the remaining useful lifetime. How do we do this? We take the historical data, from sensors for example and build a neuronal network model by relating the historical machine data with the reports about what was broken. We use machine learning to teach the neuronal network the relationship between the data from the sensors and the root causes for breaking.
In the drill down we can then see, for instance, that within the next ninety days the gearbox has a high probability to break. We use our machine learning tool to predict the time to failure (TTF) for the next sixty, ninety, or a hundred and twenty days.
PUF: Can this kind of machine learning make a big difference?