The National Infrastructure for Artificial Intelligence on the grid, NI4AI for short, is a 3-year ARPA-E initiative funded under Open Innovation 2018. Our goal is to provide the infrastructure — including data, analytics platform, and user community — to catalyze the use of AI on the grid.
- Remove obstacles to developing new AI use cases
- Host open access data to train algorithms to study specific problems on the grid
- Train analysts to work with next-generation grid data
- Build a community to exchange expertise between engineers and analysts
- Connect analysts with stakeholders in industry who could benefit from the tools they develop
AI algorithms are only as good as the data available to train them. NI4AI hosts real data provided by utility partners and sensor hosts. We want our data to help you advance your research, learn about AI and big data, develop open source software for the grid, or start a company.
Powered by PingThings’ PredictiveGrid, NI4AI gives you a high performance platform to store, access, and analyze sensor data. PredictiveGrid is a commercial tool that research groups and utilities use to work more efficiently with big data. By making this platform available to you for free, ARPA-E is providing software infrastructure that will make analysts more efficient and more productive at developing new use cases for AI on the grid.
NI4AI brings together innovators from power systems, machine learning, and computer science. Our goal is to facilitate connections between data analysts, software developers, and industry stakeholders. We share work that analysts are doing through our github, newsletter, and blog. NI4AI also provides content, tutorials, and live training to get new analysts familiar with the data, the methods, and the problems that need to be solved. Our events page lists upcoming tutorials, hackathons and kaggle competitions geared at connecting users with problems that our industry stakeholders need solved.
NI4AI is an ARPA-E initiative led by a startup called PingThings in partnership with UC Berkeley.
PingThings built the PredictiveGrid platform as a commercial software tool to streamline big data visualization and analysis workflows for utilities. The team at PingThings brings deep expertise in computer science so that analysts who aren’t computer scientists can work more readily with big data. Their core innovation is to create a novel time series database that gives unprecedented data ingestion and querying speeds. PredictiveGrid supports researchers, industry practitioners, and software developers who work with time series datasets.
The team at UC Berkeley is focused on building partnerships with analysts in research and industry to use NI4AI to modernize the grid. Berkeley is identifying and training analysts from different disciplines so that more people can work on developing new applications for AI on the grid. The Berkeley team is building educational content and tutorials so that analysts can more readily contribute to advancing state of the art, and so that the industry can more readily benefit from cutting edge tools.