I have a device that constantly fluctuates it’s power usage, which led me to thinking. It would be nice if we could have a remote device, similar to a Kill A Watt, that reports to sense about a specific devices usage.
At first glance this seems to go against the purpose of the sense solution however my device seems to ‘pollute’ the data sense is viewing. This would allow Sense to subtract out the ‘noise’ from the signal. The device could connect via power line communication or wifi whichever led to a cheaper BOM.
@CarlT,
I like your idea ! I have already questioned whether Sense might consider “connecting” via existing open APIs with other connected power sensing devices like a TP-Link Smartplug or an Elgoto Eve Energy (both of which I have been using to monitor Always On power in my household). Being an electrical engineer, I do question at what granularity a wall plug type unit might be useful given Sense device identification techniques. A few considerations:
The Sense probe samples at 4M measurements per second. Assuming that this is includes solar and that voltage and current are separate measurements x 4 coils/voltages = 500K samples second. Trying to do a sample by sample “noise” cancellation from a remote wall plug device would be nigh impossible to synchronize due to rates, distance and time domain distortion induced by the home wiring.
What about canceling things out at the level of granularity of the Sense “Power Meter” ? Most people looking at the “Power Meter” forget that its reading is a highly processed and abstracted version of the the real data. AC power, as it is delivered, is intrinsically lumpy, bopping from essentially 0W to a current driven peak every 120th of a second, like this:
The “Power Meter” is likely tracking an average based on the integrated time window of this waveform (really Imeasured x Vmeasured). So “cancelling” at the “Power Meter” level would require quite a bit of signal processing in the remote unit, perhaps on the order of what is found in the Sense unit itself. Maybe not such a good BOM cost. And then there’s the question of whether that level of “canceling” or “identification” is useful to Sense’s identification algorithms.
I’m betting that a simple “it’s on” or an “it’s off” plus a regular half second average of RMS power would be sufficient to feed Sense with 90% of the most useful information, given what we have seen of the feedback mechanisms to their data analysis so far. Most existing connected wall plugs are capable of that, plus they would also give feedback on whether the plug is turned on or not, not just the power reading.
You should be able to sync the data if both sources use a common time provider or better yet the remote device could get it’s time from the sense unit (maybe using NTP) and then just pass the data along. Then timing would no longer be an issue.
You can get one unit and just move it around from device to device to get better detection. It’s self powered (like the kill-a-watt) and if you could let it gather data for several uses until it properly learns the device.
One additional thought. It might make sense for Sense to read what our revenue grade meters are reporting for any calibration or error checking, since the PG&E meter is the ultimate metering source for billing. My PG&E net meter talks to HAN validated devices via Zigbee. Would probably require new hardware that talks Zigbee (not sure which TI SoC Sense is using), but that additional hardware and software could give Sense a much better picture of aggregated power usage for comparison.
I agree this is a "must have’ feature. A device, that you can plug an appliance into, that will help calibrate the sense signatures. I have a large commercial fridge, what you would find in a convenience store, in my basement to store beer. It consumes a lot of power but unfortunately Sense only detects when the compressor goes on. There are fans that are continuously circulating air that consume a significant amount of power, but are reported in the “always on” bucket. Because of this Sense limitation, Sense underreports it’s power consumption by almost a third compared to a Kill-A-Watt that the fridge is plugged in to. A portable module that I could plug the fridge into, for let’s say a period of a week, that talks to Sense would increase the accuracy of the model significantly. Personally, a predicted consumption of a device that is closer to the actual consumption is more important to me as a customer than what Sense actually thinks in measures the consumption to be. To make this even easier, repurpose the Solar Panel ports and I could clamp the ammeter on the circuit that is supplying power to the fridge.
