Page 227 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

2 Use Case Description

2�1 Description

The transition to renewable energy has made weather a critical factor in grid management. Change 4.2-Climate
Cloud cover remains a primary cause of solar energy forecast inaccuracy. Open Climate Fix
developed an open-source AI system that forecasts cloud movements using satellite imagery,
improving short-term solar power prediction.

This innovation has increased solar forecast accuracy by 5% at a 2-hour horizon—double the
accuracy gain seen with traditional weather services. The AI model is now being integrated into
the Quartz Solar platform for the UK, in partnership with the National Energy System Operator
(NESO), and is expected to reduce carbon emissions by over 100,000 tonnes annually.

The system uses the simVP architecture to predict future satellite images across 11 spectral
channels and another model to convert those predictions into solar PV power forecasts. This
method outperforms current ECMWF and Met Office models and is designed to scale globally
to aid electricity grid operators, solar providers, and policymakers.

Use Case Status: Pilot integration planned for Summer 2025 in the UK

Partners

Alan Turing Institute – Research collaboration during the early development phase.

2�2 Benefits of use case

• Improved forecasts support solar energy integration and smart grid management.
• Demonstrates cutting-edge AI and infrastructure support for smart energy systems.
• Reduces reliance on fossil fuels and supports grid decarbonization efforts.

2�3 Future Work

The development roadmap includes:
• Stage 1: Launch in the UK with Quartz Solar (Summer 2025)
• Stage 2: Expand to India and the Netherlands (2025–2026)
• Stage 3: Build and release a global, open-source cloudcasting service

Target adoption strategy includes:

• Large grids with forecasting needs (e.g., US, Japan, India)
• High-solar penetration markets (e.g., Brazil, Southeast Asia)
• Emerging and decentralized grids (e.g., Sub-Saharan Africa)

Deployment will combine direct services and open-source tools to foster global innovation.

3 Use Case Requirements

• REQ-01: The AI model must accurately predict future satellite images across 11 spectral
channels.

191