Transmissions operators are targeting biodiversity net gain to create a positive environmental legacy, but barriers to infrastructure development remain.
As part of its infrastructure development activities, one TO not only mitigates biodiversity loss but creates a gain by protecting ecosystems. But the Clean Power 2030 target would look more achievable – and sustainable – if the business was able to combine the efforts of all its teams across multiple projects and engage with stakeholders such as environmental groups early. “We are asking how we bring all of that together and find a collective way of working using AI,” says the digital manager.
Part of his plan involves the development of a constantly updated digital twin plotting routes of transmission networks, substations, overhead lines and underground cables.
We need to build quicker. We need to build better. Ultimately what we want to do is reduce the time it takes to make decisions.
This also includes building information modelling (BIM) and GIS data. “We can pull in information from the Forestry Commission about the impact on the land and data on carbon capture and display the results publicly.
“We need to build quicker. We need to build better. Ultimately what we want to do is reduce the time it takes to make decisions.”
It’s now possible to design substations using AI tools combined with CAE software.
In the US, one supplier is working with Autodesk Forma technology and an AI-powered design system specifically dedicated to substation development. And here in the UK, one of the TOs is already exploring the benefits of the software.
Autodesk Forma provides robust tools for proposal layouts using 2D site mapping and environmental analysis such as shade and wind assessments. Using generative AI technology, Forma’s capabilities are enhanced by interactive AI-driven analysis tailored to substation planning.
The system allows users to efficiently generate site-specific models during site analysis. It’s a form of ‘agentic AI’, a system that can pursue goals, plan and make decisions (like a human employee or agent) without human intervention.
Multiple specialised agents evaluate and compare one or more substation locations, generating precise data-driven recommendations, and supporting specialised tasks such as peatland assessments and carbon footprint analysis. Working in coordination with a primary orchestration agent, the final output is a comprehensive report formatted using a configurable template tailored to requirements of the project.
The SBS system is a retrieval augmented generation (RAG) type rather than a large language model (LLM). LLMs are powerful but better at generating answers based on general knowledge rather than specialised information. Instead of relying on a model’s pretrained knowledge, RAG pulls real-time relevant data in, ensuring responses are grounded in facts. Instead of a single AI making assumptions, specialised agents handle different tasks, such as analysing GIS data, regulatory codes, and financial and environmental impacts. The result is faster (and more reliable) AI-powered decision-making.
The agentic AI acts like a digital project manager, autonomously orchestrating multiple tasks and refining its assessments. Finally Subs GPT gives a clear recommendation on which site is best for constructing the substation complete with a visual report in Autodesk Forma.
Being able to accurately forecast impact and mitigations right at the start means we can make better decisions. The system could potentially even drive planning reform.”
“It’s really creating a draft of a design that an engineer or project manager can come back in and develop,” explains an engineering manager at the TO trying out the system. “As the design progresses it can analyse nature impact, carbon impact, cost and everything else. Before we get to that point, we have to put all the data and intelligence into each agent, which is the knowledge of our most experienced people.
“We also have to agree a hierarchy of decision-making and ultimately get to a point where that hierarchy is consulting with engineers, project managers and key stakeholders like environmental charities.
“But being able to accurately forecast impact and mitigations right at the start means we can make better decisions. The system could potentially even drive planning reform.”
This type of design tool supports the jobs of engineers, rather than replacing them.
The digital transformation manager at the TO points out that “AI does not replace human interventions but gives you a list of options. AI is there to support jobs, not replace them. It does things quickly without having to make a phone call.”
In fact, digital technology is a small part of solving a much bigger problem in terms of the resources available for infrastructure deployment. Or as the digital expert at a leading manufacturer of transformers notes, “AI and digital technology is just one part of it. We know we need to build more factories and recruit. We need to put billions into additional manufacturing capacity. Capturing knowledge from experienced engineers is also vital.”
And engineers themselves are in short supply. The TO engineering manager says AI has the potential to “reduce the stress on recruitment”. “There aren’t enough people to deliver the projects, so AI makes them feasible.” It’s not just engineers that are in demand, they add. “Biodiversity and community engagement teams are growing because it is such a massive job.”
AI isn’t trusted 100% without people in the loop. A director at an infrastructure consulting firm points out that “recording decisions is a critical part of the consent application process. If you can’t demonstrate how you have come up with those decisions, then you are not going to get consent.
We know we need to build more factories and recruit. We need to put billions into additional manufacturing capacity. Capturing knowledge from experienced engineers is also vital.
When you submit a planning application you must demonstrate what bits have been done by AI, and what bits have been done in a traditional way. “The implication is that people are not yet completely comfortable with the technology.”
A roundtable attendee from a regulator adds: “From a regulator’s point of view, how you reach a decision, how you got to that point, is important – that’s where explainability comes in. With AI, it’s like we’ve got a black box. There’s inputs going in and an output, and we can’t see what is being done in between.
“So perhaps there needs to be some kind of signposting. We’ve also got to think about the use case, and whether it could result in big consequences, like a power cut. If the risk is low and the AI gets it wrong, it doesn’t really matter.”
Transmission operators, the supply chain, and other stakeholders must think about gaps in data and where there are nuances to consider in decision-making.
“You need to be clear about data gaps and where the uncertainties are in the stuff you feed in [to AI].” That’s the view of the head of planning at a conservation organisation. “We don’t have everything digitised because we don’t know everything about each of our sites. The level of information we have is always mixed. And then there’s the expertise needed to understand what everything means and perform a balancing act.
“Is a bat more important than a newt? Biodiversity net gain measures one aspect of nature, wildlife and the environment. You need to make sure you don’t miss things not captured by one metric.”
The infrastructure consultant reinforces the point: “There are different objectives; are you trying to get planning permission, or are you trying to maximise your biodiversity? Are you trying to meet wider sustainability targets? Depending on which one you are trying to achieve, things are slightly different.”
Another panellist notes: “There is still work to do even around structured datasets like GIS and BIM data. It’s not always perfect, and so you need people to work on it before it can be fed into AI. Then you’ve got other non-structured data that is not really data as such, like PDFs. And if you work in engineering, there is tonnes of historical data that would be really good to feed into these processes.”
The digital manager from the TO believes that this issue is “industry-wide: water has got it, gas has got it. Data didn’t matter before. I’d love it if AI could go in and scan our existing systems and try and pull the data in. We’ve got the systems sitting there as archives, but nobody has the knowledge to go and scan it.”
…and there are other foundations to get in place that will help speed up infrastructure deployment.
“I would like to see work on a standardised library of 3D models for substation design, just so we can get that consistency. There is so much work we can build on there,” points out the CAD specialist from a TO.
Another attendee adds: “I think we have a lot of work to do to enable AI in the sector. There’s a lot of data standardisation required around existing assets, and we also must truly understand the use cases for AI – nail those down, and what the priorities are.”
“We know we need to build more factories and recruit. We need to put billions into additional manufacturing capacity. Capturing knowledge from experienced engineers is also vital.
There must also be greater confidence in the regulatory community about the use of AI in developing infrastructure.
But in the end, these early forays into the potential of generative AI for engineering are likely to followed by a normalisation of the technology “where it becomes another tool like CAD or BIM”, says one industry expert. Confidence in the results AI provides is also increasing as the technology matures.
In the meantime, the transmission industry needs to be share best practice, collaborate, discuss the many issues raised by events like the roundtable, and get the right governance and policies in place to manage the risks.
“The risk side needs to be worked through – despite the tremendous opportunity,” cautions our panellist from the regulator.
