Our team has been investigating opportunities to get compute cycles when they are cheapest, either through an innovative sensor system that watches clouds, or a clever scraping of PUC utility websites to ramp compute resources up when electricity is inexpensive.
This summer we are testing a system that will allow participants in the trial effort to attach compute resources on-supply and often hands-off (asynchronous, asymmetric, and automatic). If you have a job you want to run, and you are not in any hurry, you will get it for “free” while the sun is shining.
We have a number of subprototypes that we plan on refining over the course of the summer. These contributions include;
Single conductor for power and data
Efficient MPPT tracking and intermittent power capture
Commodity focused DIY documentation for community use
We feel this is best placed under the Automation umbrella since we have photodiodes that sense when clouds are coming and save state accordingly. We do have batteries and are additionally scraping the local hyperprecise weather websites but we are looking at making this opportunistic and very inexpensive.
We have finished phase 1 of our effort, which is the tech demo / prototype. Now we want to improve our design and integrate it into a single self-contained system. We will be using the NVidia Jetsons in order to be able to use the predictive analytics, machine learning and AI capabilities of that platform.
ABOUT THE ENTRANT
Name: Kurt Keville
Type of entry: team
Software used for this entry:
COMSOL,Maple, NVidia Jetpack suite of tools
Patent status: none