How power companies predict power outages before they happen

How the sensors predict power outages before they happen

A sensor clipped onto a power line can now predict a power outage before it happens, not just report one after it goes out. According to StateTech Magazine, ultraviolet cameras catch the corona effect, the faint light electricity gives off when it starts arcing across a damaged connection before it turns into a fire, and infrared cameras spot a transformer, equipment that steps electricity down to a voltage a home can use, heating up before it fails. Software then compares what the sensors see against a library of past faults, and decides whether to text, call, or dispatch a crew. Catching a fault before it causes an outage instead of responding after one is what utilities and vendors call predictive maintenance.

Consumers Energy already saved more time in 3 months than in all of 2023

Consumers Energy, a utility in Michigan, says line sensors on its grid saved 95,160 customer outage minutes, the added total of every minute customers spent without power, from January 1 through March 24, 2024. That already passed the 80,198 minutes the same sensors saved across all of 2023, about 18.7% more in a single quarter than the company reported for a full prior year, a comparison this article calculates from the 2 totals Consumers Energy reports rather than a figure the company states itself. The utility invested nearly 24 million dollars in smart grid technology in 2024, adding about 3,000 line sensors, its most in a single year.

Each sensor leaves a fingerprint of what went wrong, and then we can send somebody out immediately into the field to fix the issue.

Bharat Goel, the electric engineer who oversees the line sensor program at Consumers Energy. Source 3.

A decade earlier, a bigger utility ran the trial that proved the case

Pacific Gas and Electric, which serves 5.4 million electric customers, ran what its own case study calls the most comprehensive line sensor test the industry had undertaken, from 2014 to 2016. The utility deployed almost 1,000 sensors across its power lines and cut customer minutes interrupted, the added total of every minute power was out across everyone affected, by 12.6% from saved patrol time alone, by 18.0% once remote controlled switching was added, and by 19.4% once the sensor data was combined with calculating exactly where a fault sat on the line. This was strong enough that the utility folded the first rollout into its next rate case, the regulatory filing that sets what customers pay, with deployment beginning in 2017.

2 software vendors report even bigger numbers, without naming the utility

Sentient Energy, which also supplies the analytics behind the Consumers Energy program, says a separate, unnamed utility identified and prevented over 700 customer interruptions in 2024 using its Ample Insights product, representing over 160,000 minutes of customer minutes interrupted prevented.

We take a pragmatic, iterative, data-based approach to analytics, using proven technology including Machine Learning, to identify potential problems that may impact the grid.

Bahman Hoveida, chief executive of Sentient Energy. Source 1.

A separate case study published by C3 AI, a different software company, describes another unnamed utility that reports a 48% cut in transformer failures and 98% accuracy detecting a future failure before it happens. Neither vendor names the utility behind its best number, so both figures describe a real deployment that cannot be checked against an independent count.