There is a real need to modernise and digitalise the world’s power grids but investment in updating these systems has failed to keep pace with energy demands and requirements.
With the rapid pace of electrification and the ongoing energy transition toward net zero, governments and industries need to accelerate that investment. According to a new report from global technology intelligence firm ABI Research, worldwide investments in grid digitalisation are set to jump to over $150bn by 2030, up from $81bn last year.
“The benefits of the digital transformation of energy grids are huge and wide-ranging,” explained Dominique Bonte, VP End Markets and Verticals at ABI Research. “Most importantly, it enables the real-time management, orchestration, and continuous reconfiguration of increasingly complex and distributed energy networks and assets while unlocking much-needed additional generation and transmission capacity.” He added, “It also reduces costs in terms of both grid expansion and operational management, improves grid resilience in terms of reduced downtime and faster fault recovery, and enhances overall energy quality and efficiency.”
So, according to Bonte, the energy sector which is facing the new reality of both distributed and more intermittent forms of energy resources and systems, requires much greater levels of flexibility especially at the edge of the energy grid, where a static energy supply comes up against the dynamic energy demands of a rapidly electrifying environment.
Software-defined low and medium-voltage energy substations are increasingly coming into play here, in terms of facilitating flexible over-the-air functionality upgrades and configuration changes as well as managing and coordinating the 2-way flow of energy and making sure that the grid remains stable in real-time.
Along with the better management of the grid, ABI Research’s suggested that digital twins are going to have a critical role in the design, modelling, simulation, planning, and operation of energy infrastructure, while smarter metering would be needed to both monitor and provide granular edge and cloud intelligence to enable better regulation of the upstream grid in real-time.
Both AI and Gen AI are seen as playing an increasingly important role too, whether that’s in delivering preventive maintenance or demand-response orchestration as well as providing compliance reporting.
The digitalisation of the grid, however, faces numerous barriers and inhibiting factors ranging from a lack of financing, regulation, restricted ‘digital’ expertise among employees, long infrastructure lifecycles, and cybersecurity concerns.
“Going forward, it will be critical for energy utilities and technology providers to develop agile design and deployment practices, tap into innovative funding mechanisms, leverage open platforms and ecosystem cooperation, and address the human factor of embedding technology into company processes and culture,” explained Bonte.
Power demand
The conundrum when it comes to AI is that while it will have an increasingly important role in delivering and supporting this digitalised infrastructure, the amount of AI computing power that’s required – not just to enable this but to support its growing use across all sorts of industries and markets - means that electricity hungry AI datacentres are actually undermining another important government target i.e. the creation of a much cleaner power system.
Perhaps that’s now less important in the US, following the election of President Trump, but here in Europe it’s a real challenge.
The UK government recently announced a major investment in AI datacentres, but the enormous energy demands of AI have certainly raised concerns that the government could end up derailing its own clean power pledge, that is to see the UK, by 2030, doubling its onshore wind, tripling its solar power capacity and quadrupling its offshore wind capabilities.
Last year the UK shut its last coal mine, and the government has promised that by the end of the decade it will only use gas plants ‘sparingly’.
“All our modelling suggests that on the current trajectory the government’s [2030] target will be extremely stretching,” warned Kate Mulvany, a principal consultant at Cornwall Insight, an energy advisory body. “We cannot see how these targets can be met with current schemes and policies.”
Building and running the servers that form the heart of AI is hugely energy-intensive, with ever more power needed to train and run what are increasingly complex AI models. Globally, the electricity consumed by datacentres is expected to match the entire power consumption of Japan within the next two years, according to figures provided by the International Energy Agency.
According to Mulvany always-on energy is needed to support datacentres, to match their continuous running hours.
But while there are emerging energy storage technologies that could make it possible for wind and solar power to play a role in the future, in the near-term most markets where there is a large density of datacentres, they’ll in all likelihood end up using more gas power.
“It would absolutely pile pressure on already very difficult clean power targets,” suggested Malvany.
In the US, former President Joe Biden announced federal support to address massive energy needs for fast-growing advanced artificial intelligence data centres and called for federal sites owned by Defense and Energy departments to host gigawatt-scale AI data centre and new clean power facilities.
According to the then White House technology adviser Tarun Chhabra while it was vital for the US to ensure that the AI industry was able to build the necessary infrastructure the demand for electricity was only going to go one way with demand increasing rapidly - AI developers could be seeking to operate data centres with as much as five gigawatts of capacity for training AI models.
Speaking at Davos last month President Trump said that the US needed to double its energy production, in part due to the need to fuel artificial intelligence and he said that he would look to fast-track the approvals for new power plants, which companies will be able to locate next to their plants.
Trump went on to declare that companies would be able to fuel new power stations with anything they wanted, including “good, clean coal” - so much for the environment.
In the UK the government is aiming to create AI zones in which fast-track planning and infrastructure upgrades will be enabled to accelerate the rollout of ‘clean’ energy that will be needed.
But before that can happen there needs to be a massive investment in the country’s energy infrastructure so as to connect new energy projects to the grid and AI and new datacentres will only add to this load.
According to the UK’s National Energy System Operator (Neso) its 2030 clean power plans already assume a sharp fourfold increase in the energy demand required by AI datacentres by the end of the decade and the UK is setting up an AI energy council to better understand the energy demands associated with its planned expansion of AI, with clean, renewable energy solutions at its heart.
Revolutionising the energy grid
NREL researchers have been examining ways to use generative AI to revolutionise the power grid by improving decision support, predictive planning and control.
According to researchers AI has real transformative potential in terms of supporting more rapid implementation of clean energy solutions, protecting the critical grid infrastructure, and reducing both the capital and operational expenses that are associated with advanced energy technologies.
Generative AI is being trained to provide reliable information and decision-making support for applications within power systems.
Crucially, Generative AI, together with next-generation AI foundation models, is seen as enabling the provision of proactive decision support and predictive online control to improve efficiency, reliability, and resilience and will completely revolutionise the performance and capabilities of the grid network.
Not only that AI can be used to provide a more cyber-resilient grid but also one that can significantly reduce the impact of other hazards from blackouts and brownouts to ensuring that communities have access to affordable, reliable, and clean electricity.
Developing, planning and delivering future power grids will be vital, and AI can again really help with this. Planning can be radically improved by using AI to provide much faster and more efficient models, high-fidelity scenarios, and stochastic optimisation schemes for large-scale integrated energy systems.
Recently, the UK’s first smart substation was installed in Maidstone, Kent, as part of UK Power Networks’ Constellation trial. This is an initiative designed to enhance the capabilities of existing power infrastructure and accelerate the transition to Net Zero carbon emissions.
The substation has been equipped with advanced computing technology and communication systems that enable dynamic interaction with other sites, so by analysing power flows in real-time it’s possible to redirect energy where it’s needed.
The trial aims to optimise the electricity network and provide additional capacity for distributed energy generators and at its heart is AI and machine learning, allowing the network to operate more efficiently and safely. Potentially, these technologies could release up to 50% more capacity on to the grid making it easier for renewable energy sources to connect and contribute.
“This is a groundbreaking innovation for our network and the first of a series of smart substations. Enhancing the service provided to our energy-generating customers and making our network more resilient will prepare us for a decarbonised future,” said Luca Grella, Head of Innovation at UK Power Networks
The Constellation trial includes plans for five additional smart substations across the South East of England and if it proves successful could dynamically manage power settings, unlocking further capacity and supporting the integration of renewable energy at scale.
Open source
Given that the challenges of modernising the power infrastructure are immense how do we go about unlocking the potential of AI?
In a report from LF Energy it was suggested that with AI emerging as a key enabler for optimising energy systems open source innovation would be essential to accelerate AI adoption in the energy industry if it was to deliver on its promises.
While AI holds great promise and potential to help transform power systems and optimise their operation the report said that a different collaboration model was going to be necessary pointing to how open source was successfully being used and deployed in other sectors.
To deliver a more efficient and dynamic power network vendors will need to work with utilities and system operators as well as research laboratories and academia to develop the AI solutions that will be needed. They will need access to data and operational power system expertise from utilities and system operators, and open source is a far more effective way of structuring and scaling up this kind of digital collaboration.
The report said that critical applications of AI in power systems will require transparent, secure, and auditable models, which utilities and operators will most likely run and control themselves, with most of the data remaining behind firewalls to comply with privacy and critical infrastructure protection (CIP) laws and regulations.
Energy LF argues that open source fosters transparency, security, and trust, and will help to streamline meeting these requirements and those of AI regulations. Likewise, open source tools and community specifications can drive forward standardisation, interoperability and data exchange at a much faster pace, which will be instrumental in enabling and deploying AI.
In conclusion, while AI is among the biggest drivers of increased energy demand it has a critical role to play in modernising and delivering the power infrastructure that’s needed.
But as LF Energy said it’s not just the technology that will play a critical role. The energy industry will need to apply best practices from other sectors and leverage open source strategically, if it is to deliver on the promises and potential of AI in terms of speed and scale.
