AI to detect abnormalities in substation equipment

Two experts have announced that they will begin installing a solution equipped with artificial intelligence (AI) for visual/sound analysis to detect and diagnose abnormalities in distribution substation equipment of Japanese grid firm, Tepco PG. By installing the solution in 1,300 distribution substations during 2019, Tepco PG expects to reduce time spent on inspection patrols in the field by more than 50%.

Image/video-analysis AI, codeveloped by NTT Data and Tepco PG, will learn substation operational data to detect abnormalities such as oil leaks in oil-immersed transformers and damage to substation facilities, including barrier fences. The technology will also be able to read analogue meters, while sound-detecting AI will detect abnormalities such as damaged or worn bearings.

Tepco PG, like many other grid companies in Japan, is faced with the challenge of managing ageing facilities at a time when available labour is declining. The two companies decided to install the solution after its efficiency was confirmed in demonstration tests in 2017.

NTT Data’s Digital Maintenance Solution integrates processes for facility maintenance planning and implementation. Image/video-analysis AI, which combines deep-learning and video-analysis technologies, is based on the Amy Insight AI solution of Automagi, which humans normally perform, to detect abnormalities, identify objects and diagnose damage.

The sound-analysis AI learns normal sounds using NTT Data’s Monone solution. Monone visualises and analyses sound using abnormal-sound-detection technology, to detect signs of wear/damage, optimise maintenance costs and raise uptime.

Tepco PG aims to leverage its collaboration with NTT Data to further advance its power equipment maintenance technologies, reduce wheeling charge and enhance power-supply stability. NTT Data will continue enhancing the solution by integrating various analytical models and robotics technologies.