AI, machine learning drive predictive maintenance in power sector

Digital twins and AR are improving asset monitoring and field maintenance.

Artificial intelligence (AI) and machine learning (ML) are becoming increasingly important to predictive maintenance (PdM) in the power sector, helping utilities detect early signs of equipment deterioration and address potential failures before they escalate, according to GlobalData.

GlobalData’s latest report, “Strategic Intelligence: Predictive Maintenance in Power (2026),” said power companies including Ørsted, Florida Power & Light and National Grid are combining high-frequency sensor data, inspection imagery and operational history to detect anomalies, predict failure risks and improve maintenance planning and outage scheduling.

“Energy tracking is emerging as a critical reliability metric in PdM,” said Rehaan Shiledar, power analyst at GlobalData. “This helps to spot performance decline long before equipment trips or fails.”

“By translating technical condition signals into expected energy loss under forecast demand, weather, and dispatch, it sharpens maintenance prioritization around risk-to-deliver and real economic impact, particularly where revenues and downtime costs vary by market conditions and time,” Shiledar added.

Utilities are also incorporating advanced metering infrastructure and grid-sensing data into asset health and reliability programmes.

Duke Energy and Southern California Edison, for example, are using load and voltage tracking to monitor transformer and feeder stress, support replacement decisions and improve system performance.

Digital twins and augmented reality (AR) are also being deployed together to provide real-time information to field technicians.

GlobalData said GE Vernova uses digital twins for large-scale generation equipment, including turbines and boilers, alongside wearable AR and immersive headset guidance. Siemens is similarly combining digital twin technology with AR to connect physical and virtual assets and support operational decision-making.

Carbon pricing is another factor supporting PdM adoption, as equipment inefficiencies can increase fuel consumption, emissions and operating costs, Shiledar said.

“As equipment degrades through fouling, seal leakage, blade wear, control drift, insulation aging, or rising transformer losses, power plants often consume more fuel per megawatt-hour,” he noted.

Carbon pricing can further raise the cost of these inefficiencies through higher CO₂ charges.

PdM can also help reduce forced outages, trips and restarts, which can increase emissions and require higher-emitting backup generation.

Utilities and power generators are increasingly scaling PdM as they seek to improve the reliability of ageing assets whilst controlling operations and maintenance costs, Shiledar said.

The adoption of industrial internet of things sensors, edge computing and analytics is making condition monitoring more practical and scalable, while the expansion of distributed wind and solar generation is increasing the need for remote monitoring.

Rapid renewable growth is also strengthening the case for PdM, as distributed wind and solar fleets make downtime more costly and remote monitoring increasingly important.

“Safety, regulatory, and ESG expectations combined with improved cybersecurity and proven ROI, will push organizations to scale PdM from pilots to fleet-wide programs,” Shiledar said.