The home is an electrically dynamic place. Devices routinely get added, removed, and moved around within your home. On top of that, the electricity feeding your home can change in terms of power quality. All of these factors can affect what Sense sees and thus the continued accuracy of pre-existing device detections. With this in mind, even after Sense detects a device, it still must remain constantly vigilant and continue to update its device models. A common misconception is that Sense simply compares what it sees in your home to a stable “dictionary” of device waveforms. We wish it were so simple!
With the dynamism of the average home in mind, we want to share some possible reasons why Sense might lose a previously good detection.
Sense may initially discover two modes of an appliance as separate devices, say the toaster on “dark” vs “light” mode, or the different modes of the clothes washer motor plotted above. With only a few examples of an appliance, it’s hard to extrapolate the complete pattern. As we record more data and improve the detection model, the two clusters of patterns converge and so do the multiple models. They become duplicates or near duplicates. As a user, you may experience this in many ways: One device may no longer trigger at all. One device may trigger some of the time and the other the rest. One device may flicker on briefly, then the other device turn on. In cases of exact duplicates, if you delete one of these devices, the pattern remains explained by the other device (so feel free to do so!). If you’re experiencing “misfires” where Sense doesn’t consistently recognize the device as turning on, it should improve as it acquires more data.
The “mobile” device
You may also lose a previously good detection if a device is moved to a different leg/phase of your electrical panel. This will mainly happen with “mobile” devices like vacuums and clothes irons. Luckily, most devices in a home aren’t moved around too frequently. However, if a device is moved to a new semi-permanent location, Sense should develop a new model for it in time.
Obstructed by noise
We mention “noise” quite a bit. This can also affect a previously good detection. There’s many types of electrical noise, but for Sense, the major factor is another similar device overshadowing the current one during runtime — like, a 10kW heater running concurrently with your 150W iron. Sense can usually pick apart concurrent devices, but occasionally factors arise that make this difficult. It’s especially common with rapidly cycling, high-wattage devices or variable-load devices. Noise can also come directly from your utility lines, more commonly referred to as power quality. Changes in your power quality can affect Sense’s device detection algorithms.
Learning run amok
Sometimes the machine learning process itself can run awry. As the learning algorithm incrementally improves a model, it may get stuck in a rut where small changes to the model no longer improve it (what we call a “local maxima” in computer science). The model might only match part of the pattern. Or it might trigger on another device too, but Sense may fail to determine which to split out. In other cases, especially with more chaotic device signatures, the algorithms may not initially pick the best point from which to build the device model. As the device runs again, slight differences in the data can negatively affect detection. In cases like this, it can be wise to delete the device. By deleting the device, you’ll start device detection with a different set of initial examples and may net a better model. To learn what happens behind the scenes when a device is deleted, check out this blog post.
A device for every season
Seasonality can also affect continued detection. If a device hasn’t run for, say, six months, the original model may no longer match with what Sense is currently seeing. Your home has likely changed during that gap in time, thus requiring a new model to be developed when the device is used again. Remember, your Sense monitor (and our awesome data scientists!) remain constantly vigilant and continually refine existing models based on changes in your home. Once Sense sees enough cycles of the “new” device, a new model should automatically get developed for it. We’ve been able to successfully mitigate this issue with many AC devices and are working hard to improve it with these and other common seasonal devices.
In some rare cases, extension cords or power strips can introduce enough resistance to alter a device’s waveform and thus prevent a detection via the currently trained model. But this truly is rare and really only applies to extension cords of a significant length and not the standard 3-foot power strip.
For all these points, it’s worth keeping in mind that a home is a dynamic place. The electrical signatures of your devices can change for a litany of reasons and all of those reasons can affect Sense’s persistent detection of them. The good news is that we’ve worked to make most of these occurrences rare and are working constantly to improve the accuracy and continued reliability of device detection. If you’re consistently seeing issues with device detection, be sure to let the Support team know.