Roving Flex Sensors and Device Detection

I just now ordered a set of Flex sensors and am looking forward to installing them.

I am thinking about installing the flex sensors initially on breaker #1, waiting for a week or two, and then moving it to breaker #2, and so on until I have sensed each breaker individually. The goal of this approach is that Sense will have more information it can use to detect individual devices by capturing precise usage data isolated to an individual breaker circuit. Also, this can help to isolate “always on” power to individual circuits.

Does this idea seem sensible? Will the Sense system be able to resolve individual devices when it is provided with usage data from two overlapping sources (e.g. the main and the branch)? Will a device newly detected by a dedicated Flex sensor remain in the (integrated) model when the Flex sensor is reassigned?

I don’t believe this is the intent of the Dedicated Circuit Monitoring.

The idea is that you are using them on a circuit breaker that only has 1 device on it. So a 220v water heater, 120v septic pump, dishwasher. etc…
If you are monitoring a circuit that has multiple devices on it, you will just get one bubble for everything on that circuit. If you are familiar with the Smart p;ug integration, think of this like that, but with clamps in the panel vs plugs at your outlets. If you use a smart plug with your TV, you get a bubble called TV. If you use a smart plug with a powerstrip for your entertainment center, you get a bubble called Entertainment center.
Similar to smart plug program Sense will stop trying to detect what is behind the CT because you have already told it.

If you put it on your hot water heater for a month, and then move it to your electric oven for another month, you will lose the detection of the hot water heater when you move the clamp.
I do believe that Sense will be using the data collected from all these people adding dedicated CT’s to help feed their ML system, but you are not specifically training your Sense.
Maybe in the future that will be a feature that you can use the Flex CT’s to train specifically for your home, but at the moment its just about helping identify a device that is important to you to track that Sense can’t.

3 Likes

@ben is correct; the roving idea will not work.

This feature is designed to monitor circuits that are dedicated to individual loads and requires that those loads are monitored with CTs persistently. it works quite similarly to our current smart plug integration where the data collected is used by our data science team to improve disaggregation for everybody over time. It will explicitly not result in immediate or even-near term machine learning detection of connected loads.

Drawing an analogy between smart plugs and Dedicated circuit monitoring is limiting because the circuit monitoring provides microsecond data whereas the smart plugs provide 2 second data. This additional data availability opens up the possibility to develop the capability I described.

I don’t believe the analogy is limiting. As Ryan said, just like Smart Plugs, the data collected by the DCM will be used by the data science team. The disaggregation is what is necessary for the capabilities you are suggesting. Yes, the data collected from the Flex CT’s is of higher resolution / frequency than the Smart plugs, but its still data all the same.

3 Likes