Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EPT-1.5 Technical Report

Roberto Molinaro, Jordan Dane Daubinet, Alexander Jakob Dautel, Andreas Schlueter, Alex Grigoryev, Nikoo Ekhtiari, Bas Steunebrink, Kevin Thiart, Roan John Song, Henry Martin, Leonie Wagner, Andrea Giussani, Marvin Vincent Gabler

TL;DR

EPT-1.5 shows remarkable performance in predicting energy-relevant variables, particularly 10m&100m wind speed and solar radiation, and outperforms existing AI weather models like GraphCast, FuXi, and Pangu-Weather.

Abstract

We announce the release of EPT-1.5, the latest iteration in our Earth Physics Transformer (EPT) family of foundation AI earth system models. EPT-1.5 demonstrates substantial improvements over its predecessor, EPT-1. Built specifically for the European energy industry, EPT-1.5 shows remarkable performance in predicting energy-relevant variables, particularly 10m & 100m wind speed and solar radiation. Especially in wind prediction, it outperforms existing AI weather models like GraphCast, FuXi, and Pangu-Weather, as well as the leading numerical weather model, IFS HRES by the European Centre for Medium-Range Weather Forecasts (ECMWF), setting a new state of the art.

EPT-1.5 Technical Report

TL;DR

EPT-1.5 shows remarkable performance in predicting energy-relevant variables, particularly 10m&100m wind speed and solar radiation, and outperforms existing AI weather models like GraphCast, FuXi, and Pangu-Weather.

Abstract

We announce the release of EPT-1.5, the latest iteration in our Earth Physics Transformer (EPT) family of foundation AI earth system models. EPT-1.5 demonstrates substantial improvements over its predecessor, EPT-1. Built specifically for the European energy industry, EPT-1.5 shows remarkable performance in predicting energy-relevant variables, particularly 10m & 100m wind speed and solar radiation. Especially in wind prediction, it outperforms existing AI weather models like GraphCast, FuXi, and Pangu-Weather, as well as the leading numerical weather model, IFS HRES by the European Centre for Medium-Range Weather Forecasts (ECMWF), setting a new state of the art.

Paper Structure

This paper contains 19 sections, 4 equations, 8 figures, 6 tables.

Table of Contents

  1. Introduction
  2. AI Weather Forecasting at Jua
  3. World's First Operational Global AI Weather Forecasting Model
  4. First Global 1 $\times$ 1 km Resolution Precipitation Forecast
  5. EPT-1.5: Model Architecture and Innovations
  6. Key Features and Innovations
  7. Operational Specifications
  8. EPT Model Variants
  9. Benchmarks Details
  10. Ground Truth Datasets
  11. Benchmarking Approach
  12. Methodology
  13. Results
  14. Grid Performance
  15. Point Observation Results (Operational Model)
  16. ...and 4 more sections

Figures (8)

  • Figure 1: Spatial distribution of the weather stations for benchmarking
  • Figure 2: Comparison of 10m wind speed skill score of EPT-1.5$\alpha$ (pink), EPT-1 (orange), GraphCast (green), Pangu (red) and FuXi (dark blue) vs. HRES (black) over Europe
  • Figure 3: Comparison of 100m wind speed skill score of EPT-1.5$\alpha$ (pink), EPT-1 (orange) vs. HRES (black) over Europe
  • Figure 4: Comparison of 2m temperature skill score of EPT-1.5$\alpha$ (pink), EPT-1 (orange), GraphCast (green), Pangu (red) and FuXi (dark blue) vs. HRES (black) over Europe
  • Figure 5: Comparison of 6-hourly surface (downward) solar radiation skill score of EPT-1.5$\alpha$ (pink), EPT-1 (orange) vs. HRES (black) over Europe
  • ...and 3 more figures