State Grid Deploys X-Ray Inspection Robots on Live 35 kV to 1000 kV Lines

State Grid Corporation of China has put into service an X-ray robot that inspects compression fittings on energised transmission lines across the full voltage range from 35 kV to 1000 kV, with no personnel climbing the tower and a single tension-fitting check taking 20 minutes.

The machine was one of 12 exhibits State Grid brought to the 2026 World Robot Conference in Beijing: seven complete robots and five core software packages, all built around grid operations. The software set collects five core algorithms including defect recognition, autonomous navigation, and an embodied-intelligence large model.

A second inspection robot detects failed insulators with a 100% detection coverage rate and resolves electromagnetic compatibility problems that constrain sensor work near live conductors.

State Grid also showed two newly developed underwater inspection robots, one cable-free unit for locating submarine cable faults and one tethered multi-sensor unit. The tethered robot combines optical, acoustic, and magnetic detection modules and supports three operating modes that can be switched according to water depth, terrain, and current. Guo Rui, chief scientist at State Grid Intelligent Technology Co., Ltd., said the two robots mark an important breakthrough in underwater inspection.

The conference ran under a human-machine symbiosis theme with an exhibition area of close to 50,000 square metres. It drew 373 Chinese and foreign companies showing more than 3,000 innovation products, alongside 30 international supporting institutions and over 500 domestic and foreign guests. For the first time, the event set up a dedicated central state-owned enterprise zone, where 48 central enterprises presented 12 high-value application scenarios.

That framing is not incidental. The State-owned Assets Supervision and Administration Commission of the State Council released a second batch of 60 strategic high-value artificial intelligence scenarios for central enterprises, published together with 70 sector-specific high-quality datasets selected as outstanding results. State Grid accounted for 4 of the selected projects.

One is the distributed photovoltaic intelligent inspection and control scenario submitted by State Grid's Shandong subsidiary. It deploys four agents covering PV anomaly identification, reverse overload cause analysis, control strategy formulation, and dynamic optimisation. The Shanghai subsidiary submitted a scenario for intelligent aggregation and dispatch of virtual power plants under the new power system, already covering non-continuous industrial loads, building air conditioning, user-side storage, and computing centres.

The two dataset entries were led by the Beijing and Zhejiang subsidiaries, jointly with State Grid's information and communication centre. The electrical work safety dataset spans 16 categories of high-frequency violation behaviours along with high-risk scenarios, and rests on a five-chain governance system running from standards through data supply, data governance, knowledge injection, and quality assessment. The electricity marketing dataset follows a lifecycle construction system described as 7 phases and 25 steps.

The standardisation of datasets matters more than any single robot. A utility that codifies 16 violation categories and a 25-step data lifecycle is building the inputs that make defect-recognition models transferable between subsidiaries, rather than bespoke pilots stranded in one province.

State Grid has set out a three-year action plan for research on artificial intelligence applied to electricity, and targets AI penetration across all its business lines and its entire workforce by the end of the 15th Five-Year Plan period.

For transmission operators elsewhere, the operational benchmark is the specific one: live-line fitting inspection at 20 minutes per unit without a linemen climb, and full detection coverage on insulator faults. Those are throughput and safety numbers that procurement teams can compare directly against manual inspection cycles.