Solar nowcasting with machine vision

Enabling worldwide solar photovoltaic (PV) nowcasting via machine vision and open data


Precisely how much solar energy is being pumped into the UK’s electricity grid at any time is not well known, even by the National Grid. This project aims to fix that, using a combination of AI (machine vision), open data and short-term forecasting.

Project aims

There are many uncoordinated sources of geodata about solar PV installations: government, crowdsourced, and machine vision derived. This project will establish a worldwide open data “clearing house” for solar PV geodata. The clearing house will reconcile data sources, and transform the data into clean datasets consumable directly by machine learning algorithms. The algorithms being used for PV out-turn prediction by regional and national network operators (National Grid) and commercial market participants. This “missing link” will be a force multiplier, reducing carbon missions by enabling new demand management and energy-trading innovations.


Generation forecasting has been identified as a high-leverage opportunity at global scale. This work's open data approach to solar PV installs will enable wider third-party uses, such as analysis to plan future PV installs, and bring broader economic benefits – and in fact was demanded in 2019 by the UK Energy Data Taskforce.

Specific applications of this work include:

  • The UK National Grid could potentially save 100,000 tonnes of CO2 per year and also see considerable cost savings;
  • Local network operators are currently hitting supply constraint issues: better understanding of PV will help reduce grid reinforcement costs;
  • Energy investors and traders will be able to reduce the risk of investing in PV, helping fund solar projects.


Researchers and collaborators