Labs feature suggestion

I recently had an episode over a period of about 2 weeks wherein an “power problem event” become more and more frequently. I looked at sense data to try and discover, and was not able. Now that it is fixed however, I made an observation that, I think could be turned into a detection feature (not device detection, but rather fault detection, hence my suggestion for Labs).

Starting with the root cause discovered today: a loose connection in the panel on a neutral terminal going to a sub-panel, where most of the house loads are. Under high-isa loads this connection was observed to be arc-ing. Tightening and applying some conductive paste solved the problem.

What I observed for 2 weeks, with increasing frequency was my UPS in my office kicking in multiple times a day, for very brief periods (I think the minimum cycle for line>battery>line is about 3-4 seconds; I could hear the clicking). Similarly a power conditioning unit in the entertainment center was tripping as well. Lights might, ever so briefly flicker. I observed the “event” counter on my UPS and just last night I had 30 some of these “events”. They were generally not severe enough to trip up other devices. Microwave and stove clock, for example, never lost their stride. It took some circuitry designed to catch out of bounds conditions to notice it.

Going back to my graphs I present one from when the problem was present:
image
and after the fix:
image

Knowing what I know now, all those very brief, but high spikes, I am sure were consistent with the arcing. Not the every instance we enough to trip things up, but still. Going back in my graphs this pattern was present during these two weeks, and not before, so that is consistent too.

The arcing must have, every so briefly, caused out of balance line voltages, which is why UPS would trip. I would suggest that a similar algorithm can be implemented in sense (Labs) and be used to alert of a possible problematic condition.

2 Likes

I too had a similar problem. We had our mast replaced due to storm damage and after about 3 years moved with the Sense monitor to a new house. When removing the CTs in the old panel, they had some superficial(?) melting on the plastic bodies indicating heat. A check of the torque on the main breaker leads found them loose, hence the heat. I suspect that they loosed on heat cycling and a less that adequate installation procedure; recent changes in code require a torque spec be used when tightening but it seems there is some learning to do on “setting” of the copper (or aluminum) wire to hold the connection.

There are two issues that came up from this: first was the higher “always on” readings due to connection heating. The second is the voltage drop at higher loads altering the inrush and device characteristics use for identification.

I’m not sure how the Sense Labs folks would work on this one, but there is certainly an opportunity to improve the safety and efficiency of the panel. Perhaps the a calibration or comparison the the utility reading to the Sense reading would be a way. I didn’t really have enough info from the interface to do the analysis myself.

1 Like