LG Refrigerator never detected


Hi, I’ve got a LG LFC24770ST refrigerator that hasn’t been detected in nearly a year. I’ve actually had two of them, as I’ve moved during the year and purchased the same model for the new house. The usage pattern is clear and obvious to the human eye, with a characteristic repeating pattern of about 50W above the always-on floor, and roughly 18 minutes on and 29 minutes off when it’s idle, such as overnight. This pattern corresponds to the compressor noise. I’ve never gotten a hit on any part of it; not the compressor, not the icemaker, not the defrost cycle.

Is this a fridge that doesn’t have a model in Sense? It’s a less popular model I’m sure, as it doesn’t have in-door water or ice.



Except for EVs, I don’t think Sense actually develops/trains models for specific product “models”. The detection models are much more generic, detecting things like fixed speed motors, variable speed motors, incandescent lighting, resistive heating elements (from furnaces to coffee makers), magnetrons (microwaves), fluorescent lighting, etc.

Once identified, Sense then attempts to classify the device, say a motor, as being part of a refrigerator, pump, fan, vacuum cleaner, garbage disposal, trash compactor, washing machine, etc. by operating characteristics (features). It does this using user-based crowd-sourced training. But no specifics (yet), on make or model.



If it hasn’t detected it in a year, then I wouldn’t hold your breath. I’d pick up a HS110 smartplug and use that to monitor the fridge. You’re guaranteed to get good data that way.



It’s interesting that the trend in this thread is increasingly about giving up on Sense’s detection and transitioning to plug by plug monitoring technologies.



I don’t know how you’re inferring that trend in this thread with two replies. Sense’s native detection capabilities are continuing to develop. The OP has a fridge with an inverter linear compressor, which is notoriously difficult to detect with Sense’s machine learning technology. For $10-20, I’m suggesting that the OP makes use of a new Sense feature that will guarantee good data for that particular device.



I agree. Why wouldn’t we just buy the plugs and forgot the sense?



Not two, dozens and dozens of similar comments



Please keep this thread on track. The OP asked about a fridge and it’s quickly gone off the rails.



In the near term, this is a good solution. In the longer term, it also helps us as it gives us solid ground truth data about that exact fridge which will help with its future detection.



This is my suggestion as well, regardless of the arguments for or against.

As has been mentioned, linear compressors are a whole different ball of wax so far as detection. I have one (also an LG fridge) and I know it’s signatures, but I now that I have mine on an HS110 and can see that signature split out on it’s own, in detail, I also have a little more respect for why it may be somewhat difficult to detect as there’s actually quite a few variances in the signature when you look at it in a more granular fashion. And unlike a regular compressor, a linear compressor often varies a lot…only running as fast (or as slow) as needed to recover the temp inside the fridge.

And lastly, as was also mentioned, by installing an HS110 at this point we are providing ground truth data to Sense so that hopefully, down the road, native detection WILL be a reality for others.



Thanks for the input and explanations. I guess I’ll investigate smart plugs. The TP-Link seems to get more mentions than the Wemo around here.



They tend to be cheaper if you get them on a deal, that’s probably why. Also the Wemo App is kinda sucky compared to the TP-Link one.



Also, the Wemo doesn’t remember its last power state when it’s powered up, and the HS110 does. That is a major drawback for a refrigerator, which would remain off after a power interruption without manual intervention or some external control/monitoring.

1 Like


Yes this is definitely a major reason to choose for the TP-Link HS110. It’s why I picked them over the Wemo ones as most of mine have things plugged in that I want to come back on if I have a power failure (fridges, internet routers, smart home hubs, etc)



I own an unfortunate amount of Wemo home automation equipment as well as TP-Link. I’m seriously considering yanking all of the Wemo hardware out and replacing it all with TP-Link equipment - the entire wemo infrastructure has been a dumpster fire for a few months now - the iPhone app is a huge laggy buggy mess and my $100 wemo dimmer controlling my living room lights keeps going offline and becoming completely non responsive to the app…and even it’s own hard reset button.

Buy the TP-Link hardware, trust me. It just works.

1 Like


Have done the same for my Samsung refrigerator after 2 years of not being detected, and now I can see it clearly. Wish I could do the same for some of my 220V appliances that haven’t been detected.

1 Like


I’ve put my refrigerator on a HS110 and now have a bubble besides Always On and Unknown. Yay!

I can already see some macro data I hadn’t realized before. The first is the 2W always-on. More interestingly, the power draw of an “on” cycle is related to the air temperature. As the house cools down at night, the peaks decrease over time to a minimum about 15% lower. This makes sense; the other change the fridge could make would be cycle time, and I already knew that was relatively constant. (22 minutes on, ~25 minutes off, when kept closed.)

In any case it’s fun having more data to gather, inspect, and analyze.

Tim McBrayer

1 Like


Although only quasi-supported by TP-Link, just so you know, HS110’s are 240V compatible up to 16A@240V. I’m using 2 of them to break out and measure our 2 240V 16A L2 EV chargers.

The problem with 240V appliances however is that most draw more than 16A, so something like an electric clothes dryer or stovetop aren’t options.

Depends on your application.

1 Like


I finally put an HS110 on my washing machine today. I really want to know usage. Six months long enough to wait. Got a great look at the consumption from the Sense data monitor for it.



Not an easy pattern to detect, when embedded in lots of others. Some similarities throughout, but not completely consistent.