Disappointed in device discovery

Sense leaves that up to you to scurry around your home and figure out.

well that is one of the most amazing things about the human brain. we suck at doing things like calculations quickly (or at all even). but we excel, in ways that computers can hardly begin to match, at recognizing patterns. Computer scientists have been working for decades trying to figure out how to get computers to accurately recognize faces. and yet we can recognize a face, not just of a person from one day to the next. but of a person 10, 20, 40 years later without having seen them. they could be wearing a different hairstyle, have grown a beard, and it can still be possible to recognize them.

Even NASA is relying on Human Eyes to detect patterns.

so, everyone needs to cut Sense a bit of a break. so far as I know and have been taught, this is still fairly advanced, frontier, technology. Even to government agencies and private corporations with bigger budgets and the smartest people.

@RyanAtSense I stuck up for you, now how about an export feature. I’ve been out of the loop for a bit so I don’t know if it has been released yet. I am forgiving of Sense’s lack of intelligence at recognizing patterns, so I would like to use mine. highlighting parts of the graph and assigning their values to devices for my own calculation purposes. :wink:

1 Like

Hourly and daily export has been available in the Sense web app for about a month.

Probably not enough resolution for visually identifying anything except high power EV chargers.

Some of the folks on the forum have used the unofficial REST API to poll Sense for power readings so they can get 1 second resolution of so, but that takes a little bit of coding.

I haven’t checked in for a couple months now, so that makes sense.

Thanks! shame about the resolution. I’ll have to check it out. apps like that (or maybe just that one) have been mentioned before, but I’m not a coder :frowning: I’ve tried. I just can’t get into it.

edit: I do not see “export”

edit2: whoops. Web app.

1 Like

@kevin1

I found it. yeah that’s disappointing. I’ll keep waiting for a real export feature. I wish I knew how to code. but unfortunately, I am at the mercy of the Sense team.

not whining or complaining. I’m sure it will come eventually. and if not, the device has still proven useful.

@Grandpa2390,

I find the export incredibly useful for looking at a few things:

Correlation between Sense Total Usage and Solar Production vs. my utility providers. Both utility providers only supply hourly export as well.

https://community.sense.com/t/does-anybody-have-any-cool-graphs-or-analytics-using-sense-data-please-share/3302

Sense Detection of AC units - My Ecobees supply 5 minute interval data. It’s clear Sense can benefit from Ecobee integration and supervised learning.

https://community.sense.com/t/sense-vs-ecobee/3409

Time of Use (TOU) Billing comparisons - helpful to look at how much I would have paid with different billing plans, with and without solar.

https://community.sense.com/t/sense-tou-calculator-for-pg-e/3472

But I’m guessing that Sense will eventually include export for minute or second level data for shorter windows of time. I can’t see downloading a year of second-level resolution data.

1 Like

I don’t have solar or TOU billing. definitely would be useful in those settings.

If I did, the issue would be that Sense doesn’t accurately detect my devices. that’s why I deleted them all. it frequently gets AC and dehumidifier confused and so forth. this would be important for TOU comparison. My dehumidifier runs off and on all day, and in New Orleans, it has to.

The air conditioner on the other hand. with TOU, I would try to run my AC hard at night and let it coast through the morning till noon or so. however long it takes to get from 72 to 83.

Of course, the issue there is that there are too many variables I think to do an accurate comparison in that regard. it is cooler at night, and the sun is down, so AC runs much more efficiently. I would run the AC down to 72 degrees which is something I’d never do during the day. So I probably wouldn’t use it for my AC comparison.

The appliances that do matter (dishwasher, dryer, washer), they don’t detect very well by sense. so I need to be able to export the data by selecting the beginning and end of the time it runs, and then calculating the energy use. it would not be perfectly accurate, but much more accurate than the device detection. if the AC turned on in the middle of a cycle, I could add that period of time as well, and then subtract the two etc.

I don’t have a smart meter (part of the reason I bought the sense). I checked it’s accuracy to my analog meter but… lol. it is vague to read, and I never know what time the meter reader comes around to check.

and for posting my usage online, it just seems much easier and more informative to take a screenshot than post an excel sheet. showing the sum of usage over a period.

I don’t need all of the spikes and everything. I would just like to be able, at this point, to pull total usage at time a, and total usage at time b, c, and d. and I’ll do the graphing myself.average it out. I don’t need to know that at 5:37 it was 10 and at 5:38 it was 10.5… or whatever.
Maybe I can already do this?

I don’t know. given all that. have I misunderstood you, or is the export feature, in its current state, useless to me?

edit: maybe that’s what I need to ask for. a usage meter that shows total usage. just as the power meter shows the power usage at any given time. a Usage meter that is a line graph (rather than a bar graph) just increasing from the beginning of the day to the end, adding up the total usages. or the week. or the month, year, billing period. all of the information is already there. is this feature already there? am I missing it?

if it’s not, why isn’t it? a line graph with a total usage by time line. for longer periods, you can scroll just as you would in the power meter. zoom in and out. just slowly adding up. if I can scroll to 10:05 AM this morning and see 283 kWh, and then scroll to 10:23 and see 283.5 kWh… that would do it for me. I’d be able to work with that. goals and usage have meters… but they aren’t detailed. they aren’t line graphs.

You can mimic this at 1 hour resolution using export and a few nice features of Pivot Tables in Excel.

  • Download a month or day
  • Create a pivot table with:
    • DateTime as the rows
    • Device Names as the column
    • Summation as kWh
    • Filter set to “Total Usage” (if you are looking at net usage, include “Solar Production” in the filter as well)
  • Once your have created the Pivot Table, add a second “kWh” to the summation
  • Rename the new column in the PivotTable as “Running Total”, “Meter Reading”, or whatever you want for accumulated energy usage.
  • Then left click on an entry in that column to change it to a running total vs. DateTime. Depending on your version of Excel, the route to “running total” selection will be different.

I created this one from my September hourly download in about 5 minutes (after looking up how to create a running total in a pivot table):

Pivot Table 1-hour data from Sep 01, 2018 to Oct 01, 2018.csv (12.1 KB)

But I need minute resolution.

When will device detection improve via IoT integration? Did I miss the webinar summary posting? I don’t have a $200 Phillips light but I do have LOTS of devices that know exactly when they are on and off. Sounds like it’s the solution for smart device detection, right? Can we ask for one manufacturer a month? Is that reasonable? Maybe NEST thermostats then WEMO then LiftMaster, Pentair, etc. in 6 months you could make a lot of user’s experiences wholely improved!

@Grandpa2390 you can use my C# API wrapper to get that data :slight_smile:

2 Likes

@frankwin.hooglander
Sweet! Thanks!

I downloaded it. but I’m not quite sure what to do with it. I remember a long time ago I found where someone had created a program to gather and record information from the Nest thermostat, and it was an app that would run on Splunk. Is this something like that?

I’m running Macintosh btw.

I’m researching this, but you will probably respond before I find the answer. Programming is not my strong suit. I wish it were.

edit8: well I’m at my wit’s end. Maybe a good night’s sleep will give me a fresh approach at REST APIs or whatever it is I need to get your software running. :frowning:

1 Like

I keep trying to compile different things, but I am getting errors that I can’t resolve. ugh.

You’ll need Visual Studio to run that.

You can also install Postman and use the attached config file to get some data that way. (just remove the extra .xml extension that I had to add to be able to attach the file here).Sense API.postman_collection.json.xml (14.6 KB)

In the Get Authentication Token request you’ll have to add your Sense email + password

I’ve got visual studio. nothing will compile. I’m getting errors about the using command being invalid.

I’m missing a dependency, but I can’t figure out what it is

You’ll probably just have to restore the Nuget packages.

1 Like

Fascinating thread, but what does this have to do with “device discovery”. Sense is supposed to be able to do that for the large market, not just the programmers among us.

Might I suggest a separate thread for people developing their own layers and apps on top of Sense?

It’s relevant to the thread.

@frankwin.hooglander Thanks for trying. But I don’t think it’s going to happen for me. I installed Nuget package extension, and tried to open it up in postman and got no results.Thanks for trying to help me, I don’t want to keep wasting your time. A person like me needs an executable. lol.

So after over a month of silence from support on my ticket opened in early October for poor auto discovery, I requested an update on the status and to be honest was quite disappointed by what came back. Back in October I was told that they could see the devices in the logs that were never being discovered, they asked for all the specific model numbers and details which I provide and then yesterday I was told that my poor Auto Discovery issues are only a “medium priority bug” and I might see something in late January. The replacement monitor has been installed for 2-1/2 months (yes I had to start all over again, Yeah!) I have a grand total of 4 devices that are not Hue lights discovered.

It seems the priority is new features vs fixing the foundations of the product that are still not working as advertised which is quite disappointing. At this point I guess I’ll just have to accept that it will likely never work any better than it is now, will always only show a portion of my power use (often double now with the Smart Plug double-counting Always On issue) and move on and never recommend this solution to anyone else.

It’s extremely disappointing that their attention has focused on these new features when so many people continue having disappointing results with the foundations (auto discovery) of the product.

I can’t speak for the exact issues you’re referencing without looking at the support tickets, but yes, some bugs receive higher priority than others. Our dev and data science teams are small and priorities get shifted around occasionally as other issues arise. Some bugs break more than others and those must receive priority. We also have to constantly assess the reach of a bug. Priority tends to be given to bugs affecting the widest set of users over isolated, device-specific issues.

I hear your complaint about new features. There’s always a balance to strike between fixing what’s current and building stuff anew. That’s not unique to us (I had some banging-my-head-on-the-table moments this weekend with some audio software that got updated without fixing certain issues first). We’re always keeping that division in mind and re-assessing what deserves resources. Still, even while we’ve been building new app features, we’re still pushing out bug fixes (see the three patches to v23 as an example, but many more happen behind the scenes that never receive release notes). All that said, discovery is more in the realm of the data science team. While the software team has been working on new app releases, the DS team is constantly focused on improving device detection. While they certainly have their hand in the app side of things, their work on the core functions of the Sense app has in no way been hindered by recent app updates. It’s worth mentioning too that fixes here do not come easy. It’s not a simple matter of the DS team flicking a backend switch and suddenly devices can start popping up. Detection fixes require a lot of attention and a lot of data. Those two things can take a fair amount of time. Support could have been more transparent about this upfront and I’ll be sure to pass your feedback along to the relevant teams here.

Lastly, in many cases, newly released features have been all about improving device detection. Philips Hue and the smart plug release immediately come to mind here. These took a ton of dev resources, but they’re giving us fantastic data to improve device detection for everybody. We already have a lot of ground truth data rolling in from smart plugs and the DS team can’t wait to get their hands on more.

Sorry for the incredibly lengthy response, but I get the frustration. Most of us here use Sense as well and deal with the same bugs. Luckily, device detection issues as bad as yours are rare, but they do happen. Homes vary and devices vary and these can really get in the way of a good device detection experience. If you’d like to discuss further, shoot me a PM.

1 Like