Company or Institution
Open Climate Fix
Energy & Natural Resources
Sustainable Development Goals (SDGs)
SDG 7: Affordable and Clean Energy
SDG 9: Industry, Innovation and Infrastructure
SDG 13: Climate Action
General description of the AI solution
The project proposed is an open-source solution for the short term forecasting of solar photovoltaic (PV) energy production (Nowcasting) to enable the faster green transition of electricity generation in both developed nations and the Global South.
The project was conceived after our experience working in deep learning (Google DeepMind & autonomous vehicles) was contrasted with working in the energy industry (National Grid, Siemens & Origin Energy) where less modern approaches are currently employed.
The solution comprises two parts: firstly the forecasting of the electricity generated for minutes to days into the future. Traditional weather forecasts are not optimised for forecasting solar irradiance and are slow to run, meaning forecasts for a few hours ahead are stale as soon as they are produced. Our AI solution for this combines 5-minutely geostationary satellite imagery, not normally used in energy forecasting, with historical solar power generation levels and weather forecasts. Using state of the art AI techniques produces forecasts of PV generation which are more accurate for short time scales and are updated every five minutes, giving real-time information.
Secondly: in many locations in the world, the location of solar PV assets is poorly known. We will use high resolution satellite imagery to identify solar panels and construct a map of solar PV for a region.
This map, when combined with the improved forecasting capabilities, allows electricity grid operators and other participants in the electricity ecosystem to understand and manage the growing volume of solar PV generation on the grid, making it cleaner, more reliable and lower cost.
Github, open data repository, prototype or working demo
HPC resources and/or Cloud Computing Services