AI-powered digital twin for the power system

Summary: impacts and findings

This project confirmed that the lack of visibility of electricity grid loads, capacities and constraints was hampering the ease and speed of planning low-carbon technologies.

To address this, the project team developed a ‘digital twin’ of the grid which uses machine learning and integrates large volumes of data in a platform to support planners and energy companies in infrastructure planning.

Project aims and approach

Achieving Net Zero emissions by 2050 requires reconfiguration of our energy system from ‘top-down’ to ‘bottom-up’, integrating low-carbon technologies (e.g. renewable generation, electric vehicle charge points and heat pumps) and new services. One of the greatest challenges in achieving this is poor visibility of the electricity grid’s loads and capacity, particularly in the ‘last mile’ low-voltage networks.

Farad.ai has developed an AI-powered digital twin of the UK’s electricity grid. This gives project developers and consultants much better visibility of energy demand, network constraints, land ownership, planning factors and local transport networks, so they can optimise site selection and reduce project lead times.

The product gives a fuller picture of the planning risk and the probability of a successful grid connection, allowing developers to make more informed decisions about infrastructure before engaging distribution network operators in costly connection surveys.

The platform has integrated machine learning algorithms that model local energy demand, and includes factors such as the usage rate of charge points.

Phase 1 of the project was called AI for Low Carbon Technology Site Optimisation. Farad.ai worked alongside users and customers to design, build and test a prototype platform. This was a skeleton digital twin of the grid, integrating over 40 million data points from 25 independent datasets including energy, transport and land, to give visibility of the extra high, high and medium voltage networks. It also provided a snapshot of selected locations across the low voltage network.

Phase 2 expanded the digital twin’s coverage of the network, integrating new data sets and increasing its accuracy to meet the needs of a wider range of users. The work included additional user research/testing, front- and back-end development, operational development and commercial planning. The team also worked on a second product, Sabre, a tool that predicts future electricity demand and supply constraints to enable policymakers, developers and consultants to improve their planning capabilities.

You can watch a detailed ‘show and tell’ presentation and demonstration of the project, recorded in June 2022:

The show and tell from Phase 1 of the project, recorded in July 2021, is also available here.

Partners

Alian Ltd t/a Farad.ai

Dates

April 2021 to May 2022 

Achievements and barriers

This project delivered a fully functional application, on a ‘software as a service’ basis, that provides an accurate map of demand at the ‘last mile’ of the network. This is now in use with customers, supporting planning for low-carbon infrastructure developments.

A key barrier to creating a comprehensive digital twin was the lack of metering of low voltage substations. The project mitigated this by modelling grid constraints at higher levels, then using machine learning and power systems knowledge to build the digital twin for the lower levels of the grid.

Next steps

Farad.ai continues to add functionality to the system and to bring clients on board.