Building the world's first foundation model of the electric grid.
The world's most important physical industry is under unprecedented stress.
We are using ML to build the 22nd century grid.
BUILT BY ENGINEERS,
SCIENTISTS & RESEARCHERS FROM
SCIENTISTS & RESEARCHERS FROM
Fewer outages, lower costs,
more energy delivered
more energy delivered
Products
One model for the physics, the assets, and the weather
Most of the industry sells one slice of the grid: weather forecasting, a power flow tool, or an asset intelligence dashboard. But the grid is a single physical system. The weather drives the load, the load stresses the assets, and the assets define the network capacity.
We combine all three into a single ML system, giving operations, planning, and market teams one place to connect every data point and model under deep uncertainty.
FORECASTING
Weather generation & demand forecasting
We have invested in some of the world’s best weather researchers to build the most accurate high-resolution forecast for generation and load. Our models pair numerical weather prediction with graph neural networks, and are verified against ground stations rather than reanalysis, where most AI forecasts lose accuracy.
Updated as new data arrives: 3 km resolution, hourly, across 48-hour, 15-day and 12-month horizons, resolved to the asset, the node, or the market.
MAPPING
Asset discovery & network mapping
Most utilities still can’t pinpoint an asset’s location or condition. We take the records you already have and rebuild the rest from satellite, drone, street view and field imagery: topology, asset condition, vegetation risk, and connectivity from source to end point.
MODELLING
Network modelling & predictive intelligence
In an ideal world every node on the grid would be metered. Almost none are, which leads most people to assume grid behaviour can’t be learned at that resolution. It can. Our graph neural networks infer what the unmetered network is doing from the parts that are measured: state estimation with partial metering, losses and overloads located, and power flow and OPF across transmission and distribution.
PLANNING
Contingency & risk planning
Most planning tools still assume a single forecast and a fixed set of contingencies, but the uncertainty envelope around weather and demand has widened. We run thousands of weather and topology scenarios against your network and show which ones actually breach a limit: N-1 and N-1-1, failure propagation under extreme weather, thermal models for lines and transformers, and dynamic line ratings.
Foundation models, designed for the grid
VISION · WEATHER · PHYSICS
Turns satellite and street imagery into a geo-referenced network graph, ready for physics.
Computer vision
Deccan
Computer vision
Deccan
High-resolution forecasts from thousands of sensors, modelling non-linear atmospheric dynamics.
Weather
Indus
Weather
Indus
Learned solvers for grid physics across contingencies, from bus-branch topology and telemetry.
Power flow
Mycelium
Power flow
Mycelium
Get started with Pravāh
Get in touch