National Grid Transmission: Using AI to see risk before it happens

How do you know which assets matter most?

The transmission network is under more pressure than ever. Demand for electricity is rising, the system is more interconnected and the weather is no longer doing what it used to. For National Grid Electricity Transmission (NGET), which owns and runs the high-voltage network, knowing where the risks sit at an asset level is critical.

Innovative thinking on how AI can support engineering expertise

FRAME is a project we’ve been delivering with NGET to help meet that challenge. It’s still at the innovation stage – but it already demonstrates what well-calibrated AI risk modelling can do for electricity transmission resilience when it’s built in genuine partnership with the engineers who run the network. Together with NGET, we’re using it as evidence in a broader conversation about how data-driven methods can support engineering expertise, not replace it.

The approach we’re using isn’t a break from how risk has traditionally been assessed, but an evolution of it. Expert judgement remains embedded in the model – what changes is its ability to capture how multiple risks combine and interact, rather than looking at each threat in isolation. The practical result is a tool that helps operators decide where to act first, not just which assets hold most risk.

Building a unified risk framework

From day one, we’ve been working closely with NGET resilience engineers and asset specialists to ensure this project will deliver a practical solution grounded in real-world knowledge and experience. 

First, we pulled together the data influencing asset condition – age, location, weather exposure, insulation type, performance history and more – and unified it into a coherent dataset. From there, we developed data-driven risk models using AI to estimate asset-level risk and improve accuracy as more data became available. We sense-checked outputs against existing engineering frameworks and, just as importantly, against expert judgement.

How this approach breaks new ground 

Traditional risk models look at one threat at a time – say, wind speed or flood depth – and treat all assets as broadly similar. But real infrastructure doesn't work that way, and neither does a changing climate, which tends to hit systems with multiple stresses simultaneously.

More sophisticated models can account for several factors at once, and can be tailored to a specific piece of infrastructure under specific conditions – asking not just ‘how bad could it get?’ but ‘how bad could it get for this substation, in this location, under this combination of conditions?’ FRAME is one example of this kind of tool in practice.

A repeatable, transparent view of network risk

The project has delivered the blueprint for a bespoke solution that NGET resilience teams will be able to use every day. It provides:

• machine learning models – including calibrated predictive models for ground assets and overhead line circuits – that bring siloed asset information into a single, data-driven view of resilience
• a rules-based framework to translate individual asset risks into substation-level evaluations
• a straightforward way to review assets and associated risk profiles in one place.

Most importantly, it has the potential to make decision-making smarter. Conversations about how to optimise grid resilience will be grounded in a shared, transparent view of risk. 

The case for what comes next 

We continue to work with NGET to develop the solution – the case for investment is compelling. The real opportunity, however, lies ahead. 

As this project moves from proof of concept into business as usual, its ability to reduce engineering worktime, improve voltage and reactive power control, and increase asset availability will compound over time. The foundation is in place. The next phase is about embedding it so that smarter, evidence-based decisions about grid resilience become the norm – not the exception. 

Want to read more on our work on FRAME? The full report is available on the ENA Smarter Networks Portal.