More data
Utilities now hold more than in earlier technology cycles.
AI, defined
NIST defines AI as a machine-based system that makes predictions, recommendations or decisions toward human-defined objectives.
Computer science
AI
Machine learning
The problem
Faster change
Network conditions shift faster and are harder to predict.
Fewer outages
Grid equipment must be available more of the time.
Flip the switch
A power line with a fault. The relay finds it, the breakers on both sides open, and the rest of the line is restored.
- 1Fault
- 2Locate
- 3Isolate
- 4Restore
Drawing only. Timings are not to scale.
Fault protection
US Department of Energy (DOE), August 2025
Traditional protection equipment
AI-based protective relaying
Can be about 100 times faster
at locating and isolating faults, DOE says.
Traditional
AI-based
At that date, DOE was planning tests with utility partners.
Chip design
IEEE Spectrum, September 2026
An optimised human baseline for the matrix multiplication units.
10% less area
for the matrix multiplication units.
The numbers
Up to 30%
Predictive maintenance could cut grid outages, in E.ON research the IEA cites.
30-50%
AI-based fault detection can shorten grid outages, the IEA says.
Around 300 TWh
terawatt-hours of electricity could be saved worldwide if existing AI-led interventions scale up.
Data centre demand will roughly double
485 TWh2025
950 TWh2030
In the UAE
DEWA Distribution Network Smart Centre
It uses Big Data, AI and machine learning for diagnostic and analytical services.
Abu Dhabi Department of Energy
Its Google Cloud agreement release mentions predictive maintenance for power plants and networks.
UAE government pilot
A 2026 pilot applies AI to optimise energy systems, cooling systems and operational loads of digital infrastructure.
AI assists engineers
- Monitor
- Forecast
- Advise
In 2026 the IEA said digital tools, AI included, can monitor, forecast and advise on limits.
An IEEE-affiliated blog, not official IEEE policy, says large language models augment engineers.