As solar energy leads the global shift toward renewables, its intermittency remains a significant challenge for energy traders, grid operators, and planners. Accurate solar forecasting is essential for balancing supply and demand, optimizing energy storage, and maintaining grid stability.
Helios is a next-generation solar irradiance model, built on Zeus AI’s EarthNet foundation model. By processing high-resolution satellite imagery in real time, Helios delivers unprecedented accuracy and speed.
Figure 1: Global Horizontal Irradiance (GHI) predicted by Helios. Helios integrates solar geometry, clear-sky modeling, and live observations to produce highly accurate solar predictions within minutes.
Superior accuracy. Helios outperforms traditional models, particularly under challenging, cloudy conditions. More detailed verification is presented in the Helios intro slide deck linked at the end of this post.
Minimal latency. Traditional numerical models can take hours to generate forecasts, which is too slow for many operational uses. When minutes matter, Helios delivers forecasts in under 15 minutes, enabling real-time decision-making.
Observations-driven approach Helios’ competitive edge comes from its innovative methodology: Live observations power the forecast rather than slow, physics-based models.
Helios is powered by the EarthNet foundation model, which uses an advanced transformer architecture to learn spatiotemporal relationships across diverse data sources. This technology allows Helios to generalize without task-specific training, performing zero-shot forecasting without explicit training.
Unlike traditional models, Helios has no reliance on Numerical Weather Prediction (NWP) models. In validation with ground measurements, Helios meets or exceeds National Solar Radiation Database (NSRDB) accuracy – without any NWP inputs.
Instead, Helios rapidly integrates data at orders of magnitude beyond the capacity of numerical weather prediction (NWP) models. Satellite observations track cloud movement, formation, and opacity in near real-time, while solar geometry and clear-sky modeling factor in seasonal, daily, and hourly variations in sunlight.
import xarray as xr f = 's3://zeus-public/helios/v1/helios-v1-historical-0.06-hourly.zarr' ds = xr.open_zarr(f, storage_options={"endpoint_url": 'https://fly.storage.tigris.dev', 'anon': True})
Figure 2: Helios delivers greater accuracy than HRRR in analysis and short-term forecasts, ensuring reliable predictions for spot markets and intra-day decision-making.
Helios V1 is just the beginning. With ongoing enhancements to resolution, coverage, and refresh rates, we’re committed to staying at the forefront of solar forecasting. Our roadmap includes:
Contact us to access real-time actuals and 6-hour forecasts.
With Helios, Zeus AI is revolutionizing solar forecasting by delivering faster, more accurate, and real-time predictions powered by EarthNet. Whether analyzing past trends or optimizing real-time decisions, Helios provides the tools to navigate the evolving energy landscape with confidence.