Advancing Solar Irradiance Prediction with NVIDIA Earth-2

Summary

Predicting solar irradiance is crucial for optimizing solar power generation and grid management. Recent advancements in machine learning and artificial intelligence have significantly improved solar irradiance forecasting. NVIDIA Earth-2, a comprehensive platform for building climate digital twins, has developed a new solar surface irradiance (SSI) model that combines numerical weather prediction (NWP) and AI to predict solar irradiance at a global scale without the need for real-time data. This article explores the NVIDIA Earth-2 platform and its capabilities in advancing solar irradiance prediction.

Advancing Solar Irradiance Prediction with NVIDIA Earth-2

Solar irradiance forecasting is essential for the efficient operation of solar power plants and grid management. Traditional methods rely on numerical weather prediction (NWP) models, which have limitations in accuracy and scalability. To address these challenges, NVIDIA, in collaboration with the Institute for Atmospheric and Climate Science of ETH Zurich, has developed a new solar surface irradiance (SSI) model using NVIDIA Earth-2, a full-stack platform designed for building climate digital twins.

NVIDIA Earth-2 Platform

NVIDIA Earth-2 is a comprehensive platform that enables fast and accurate weather and climate physics modeling, leveraging NVIDIA’s accelerated computing capabilities. It provides tools for training and inference of AI models for forecasting, downscaling, interpolation, and other applications. The platform includes NVIDIA Modulus for model training and Earth2Studio for model inference, allowing users to select from a wide range of data sources, architectures, and pretrained models.

New SSI Model

The new SSI model developed by NVIDIA combines advances in NWP and AI to form a model that is more scalable and robust than traditional approaches. This model can predict solar irradiance at a global scale for multiple days without the need for real-time data dependent on satellite coverage. The model is trained using NVIDIA tools and integrated into an Earth-2 pipeline, enabling straightforward production of high-resolution, real-time predictions crucial for applications in climate science, renewable energy, and weather-dependent industries.

Earth2Studio for Model Inference

Earth2Studio is a Python library designed for AI-driven weather model inference. It provides a comprehensive set of building blocks tailored for both deterministic and probabilistic weather predictions. The library supports a wide array of data sources, including the Climate Data Store (CDS), ARCO, the Global Forecast System (GFS), the Integrated Forecasting System (IFS), and the High-Resolution Rapid Refresh (HRRR). Users can choose from a range of pretrained models present in NVIDIA GPU Cloud (NGC) and seamlessly integrate custom models into the framework.

Running a Weather Forecast Using Earth2Studio

To run a weather forecast using Earth2Studio, users can follow a simple workflow:

1. Import necessary data sources and models: Import data sources such as IFS and models like FourCastNet SFNO and SolarRadiationAFNO from the model registry (NGC).
2. Specify custom models: Users can specify custom models trained with Modulus and added to NGC for convenience.
3. Define data source and output storage: Specify the data source, output storage, and number of timesteps.
4. Perform simulation: Call the run.diagnostic function to perform the simulation.
5. Postprocess forecasts: Use downscaling tools like CorrDiff to postprocess the forecasts.

Visualizing Weather Forecasts

The Earth-2 platform provides visualization pipelines leveraging NVIDIA Omniverse, a 3D development platform designed to facilitate interoperability between different applications. With NVIDIA RTX real-time rendering technology, Omniverse enables the creation of high-resolution, interactive visualizations ideal for a wide range of industry use cases. Users can build interactive, detailed representations of the Earth and atmosphere, facilitating better understanding and decision-making.

Example: 7-Day SSI Forecast

Figure 5 shows a 7-day SSI forecast constructed with the help of NVIDIA Omniverse. High irradiance is indicated by a bright color overlaid on top of a satellite base image. This provides a glimpse of what’s possible in terms of building a visualization pipeline for high-fidelity rendering of Earth system data.

Getting Started with NVIDIA Earth-2

To get started with NVIDIA Earth-2, users can follow these steps:

1. Set up the environment: Quickly set up the environment with the Modulus start guide.
2. Train a model: Visit NVIDIA/modulus on GitHub to see a few examples.
3. Install Earth2Studio: Follow the Earth2Studio Install Guide.
4. Run a weather forecast: Follow the NVIDIA/earth2studio GitHub tutorials to run a first weather forecast.

Table 1: Comparison of Traditional and NVIDIA Earth-2 Approaches

Table 2: Benefits of NVIDIA Earth-2 for Solar Irradiance Prediction

Conclusion

Predicting solar irradiance is crucial for optimizing solar power generation and grid management. NVIDIA Earth-2, with its new SSI model, offers a robust and scalable solution for predicting solar irradiance at a global scale without the need for real-time data. By leveraging NVIDIA’s accelerated computing capabilities and combining advances in NWP and AI, the NVIDIA Earth-2 platform provides a comprehensive toolset for training and inference of AI models, enabling accurate and high-resolution predictions. With its flexibility in model selection and seamless integration of custom models, Earth2Studio makes it easy to produce data that can be easily visualized and postprocessed. By advancing solar irradiance prediction, NVIDIA Earth-2 contributes significantly to the efficient operation of solar power plants and grid management, paving the way for a more sustainable energy future.