Case Study

A leading investor-owned utility in the Midwest recently leveraged intelligent line sensors and advanced analytics to preempt outages due to equipment failure showcasing the power of predictive analytics for reliability and resilience improvement. With multiple intelligent line sensors deployed on a feeder, one of the line sensors, located just downstream of a hydraulic recloser, captured multiple low amplitude (<10A) disturbance signatures. These disturbances were classified by Sentient Energy’s machine learning models as precursor anomalies, correlated with active equipment failure. The observed activity downstream of this sensor site resulted in the sensor site location escalating to a “Watch” status.

Approximately a week later, similar precursor anomalies resurfaced, leading to a second escalation at the sensor site, a “Warning” status. Sentient Energy engineers reviewed the escalation with the utility, drawing attention to the imminence of an equipment failure. Roughly two weeks after the initial Watch indication and with the evidence gathered over that time, the utility dispatched a crew to conduct a field investigation. The crew initially found no clear evidence of equipment failure in their exploration despite the evidence observed in the electrical disturbances.

The failing equipment did not remain hidden for long though. A few days after the field inspection, the line sensor reported a single-phase fault with a momentary interruption cleared by a hydraulic recloser operation. This event was not visible to the utility prior to the line sensor installation due to the lack of SCADA visibility at this site. Based on the fault current captured by the sensor, however, potential fault locations downstream of the device were identified. The utilities initiated a second field investigation occurred a few days after the momentary fault, where the crew identified a cracked fuse cutout at the second identified location. The cutout was replaced and the precursor anomalies for this event ceased after the repairs were made.

This example highlights how the MM3ai System, leveraging advanced electrical disturbance classification algorithms and predictive outage models, can identify an impending failure and alert the utility prior to a sustained outage. Precursor anomalies do not necessarily occur at regular intervals, but rather when specific environmental or electrical conditions are met. In this instance, significant precursor anomalies occurred on four separate days, followed by a fault. None of the precursor anomalies prior to the fault were detectable by traditional distribution protective devices given the subtleties.