MSEDCL pilots AI-based system to predict transformer failures

Chhatrapati Sambhajinagar: Seeking to end the cycle of recurring transformer failures and technical losses, the Maharashtra State Electricity Distribution Company Ltd (MSEDCL) has launched a first-of-its-kind Artificial Intelligence (AI) platform named “Vitaran Intelligence”.

The pilot project, initiated in the Chhatrapati Sambhajinagar division, aims to transform grid management from a reactive “repair-after-failure” model to a proactive preventive maintenance system.

Despite the large-scale installation of smart meters, many power utilities struggle with limited visibility into their distribution networks. Authorities said this initiative is the first in India to leverage advanced data science to solve these long-standing infrastructure challenges.

The platform was conceptualised by Aditya Jiwane, a 2021-batch IAS officer and joint managing director of MSEDCL’s Chhatrapati Sambhajinagar division. It has received the green light from MSEDCL chairman and managing director Lokesh Chandra and aligns with the Centre’s revamped distribution sector scheme (RDSS).

“The pilot combines AI, machine learning (ML), feeder analytics, and thermal stress estimation to convert raw smart meter data into actionable operational intelligence,” Jiwane said. “It helps utilities identify vulnerable transformers and feeders before a breakdown occurs, allowing officials to intervene early.”

Currently, Indian DISCOMs (distribution companies) face high technical losses, overloaded feeders, and frequent transformer burnouts. Traditionally, these issues are only addressed via field inspections or after consumer complaints are filed.

Vitaran Intelligence changes this by monitoring electricity consumption patterns and transformer loading behaviour in near real-time. Using thermal loading models inspired by international IEC 60076 standards, the system classifies transformers into four categories: healthy, moderately stressed, highly stressed, or critical. This allows engineers to prioritise maintenance where it is needed most.

One of the most innovative features of the project is the use of “agentic AI”—autonomous systems that go beyond simple dashboards. These AI agents generate automatic alerts, operational summaries, and specific maintenance recommendations for field staff.

Crucially, the platform achieves this using existing smart meter data, avoiding the need for expensive new hardware installations. Its “feeder intelligence” module also identifies specific pockets of abnormal demand and areas where infrastructure upgrades are most urgent.

Power sector experts believe this shift toward data-driven governance is a significant milestone for India’s power sector reforms. By moving to preventive maintenance, MSEDCL expects to reduce emergency repair costs, minimise downtime for consumers, and optimize its workforce.

If the pilot proves successful, officials intend to scale the AI-powered ecosystem across the state, building a foundation for future smart grids and more reliable, transparent DISCOM operations.