I’m excited about my new Sense Power Monitor and just a few minutes ago it detected it’s first device.
Just to develop a comfort level that the usage billed to me by the Utility Company and what I’m actually using I’ve started a process of utility meter readings and comparing them to the Sense usage. In a 24 hour period (top-of-hour to top-of-hour) I get a 14kWh discrepancy. Utility meter is showing 14kWh more than the Sense usage for the exact same time period.
Does anyone have a good reasonable explanation before I go to the Utility to complain of an inaccurate meter?
Hayward, welcome ! Here are a few thoughts and questions.
14kWh is a lot. What is your house configuration ? Do you have solar w Sense Solar ? How much energy do your typically use in a day ? Are you sure you have your Sense on the house mains and not just a single sub panel (yes, that has happened several times on this forum) ?
Take a look at your Power Meter and Trends to see if there are any places that look like Sense has dropped data (total usage drops to zero). There was an issue with the Sense API and backend a few days ago, that did not drop data, but left some unprocessed. The fix has been percolating through everybody’s readings, but might not hav reached you yet.
Which utility do you use and how did you get access to the utility measurements ? Many utilities use smart-meters, but that doesn’t guarantee accuracy if you are looking at their measurements via their website. Some utilities only do “actual reads” every so often, even with a smart meter in place. In between they do “Calculated reads” which can be off by a large margin, especially over a days time without any more actual reads.
Do a more detailed comparison to find the problem area. This posting includes a bunch of ideas for easiest to hardest:
Thank you for your quick reply and helpful information.
No solar just a typical home drawing power off the grid.
How much I use per day is a bit difficult to get too. It certainly depends upon the weather and here lately in my region the weather has been a bit wonky, mid teens to low single digits (Celsius). I do have a complete record of our power usage going back 8 years. I also do a comparison based upon degree day during that same period. I have weather data going back to 1942 and use it as a prediction tool for estimating purposes.
Just as an aside maybe the Sense folks can incorporate degree day analysis into their modelling as it will be a good basis to determine house energy efficiency.
I have confirmed that the Sense sensor clamps are clamped on the two mains inside the electrical panel.
Over the 24 hour period the Sense data is missing approximately 1 minute of data, where the graph drops to zero.The period I have started my observations is from Nov. 4 at 9AM local time to Nov 5 9AM local time.
The utility meter is digital mounted on the outside of the house. I have been taking readings at the top of the hour and then compare them to the Sense total energy usage for the same hour. For example these are the comparisons this past six hours (ending at 2PM local time, less than an hour ago):
Utility company reading 31 as compared to Sense 26.8. That’s 4.2kWh over the six hour period.
The meter is a Centron with a digital display manufactured by Itron USA. It is a 0.5-200A 240V 1PH TYPE C1SR 3 Fils wire. It also says on the meter
FM 2S
60Hz
Kh 1.0
Thanks again for your suggestions, I’m at a loss as to the discrepancy.
Hayward
I did a quick plot and linear regression on your hourly data. The data has reasonably good correlation, but the fit line indicates your Sense power reading is 0.8396 * Utility reading + 129Wh. But the spread is enough to suggest that something else is a foot here. Can you get your meter reading via hour from your utility website ??
Thank you kevin1 for providing the analysis. I don’t fully understand it, but I have asked the Utility company for the hourly data. Their website says that their readings are done on a monthly basis when a technician traverses the neighbourhood and pings the meters in the area. So I’m doubtful if they will provide me with hourly data.
It concerns me that there is a 14% discrepancy in the Sense data and the Utility readings. I have asked for the Utility company to help me reconcile this discrepancy.
In the meantime how can I get a .csv of the total energy usage of the Sense data. I note the ability to export specific devices but I haven’t discovered how to capture the entire energy usage.
Select the given time frame from on the Trends page, export a .csv of data (by day would probably work for this specific request.)
Drop the data into a Google Sheets document (remove the first row with the description text).
Add a filter to the data
Select “Name” column, filter by “Total Usage”
You’ll then have a daily kWh consumption csv. If you want to visualize the data a bit more, you can play around with in Google Data Studio to get a graph of the usage.
Here’s a graph of the daily kWh from my home YTD using the method above:
@hjkeats, thanks for the feedback. If your utility can only provide you with monthly data, then you are going to have to use your meter as the reference. I’m guessing your meter cycles through accumulated energy usage, and instantaneous power on the display ? If so, I would suggest two things.
Take daily snapshots of your accumulated energy (kWh) usage at midnight (a pain, I know) and compare against Sense exported data in a graph similar to @JustinAtSense ‘s or a scatter plot like mine (daily utility vs. Sense). I suggest daily since it seems like the resolution on your meter (only displays results to the nearest kWh) is very poor compared to Sense. If you want to reduce "representation error’ percentage, daily samples are better than hourly.
Other option is to compare Sense instantaneous power on the Power Meter vs. your meter (assuming that the display cycles to current power usage). Be aware that the sample interval for the meter is different from Sense (1/2 second), but when the power usage is fairly stable in your household, the number should be close, even if the sample interval is different. I suspect in this situation, the resolution of your meter will be more in line with Sense so you might get a comparison that has less representation error in it.
ps: If I had to guess from my graph, I might suggest that one or both of your Sense CTs might no be fully closed. I have seen similar (Sense less than Utility) when one of my CTs wasn’t fully closed.
Thanks once again for the input and the things to check for. If I may include a couple of pictures of the sense clamps and the electrical panel. I looked closely at the sense clamps and they are as tight as I can get them.
I have a spreadsheet which I’m tracking the hourly data with the observed readings from the utility meter. If you wish I can send you the spreadsheet. I plan on performing these observances for the next foreseeable future until the differences are resolved.
The first twenty four hours 9AM 4 November to 9AM 5 November the difference was 14kWh on a total usage of 99kWh in accordance with the utility and 85.1kWh as identified by Sense.
The second 24 hours from 9AM on 5 November to 9AM on 6 November the difference was 14.7kWh based upon power consumption in accordance with the utility meter of 98kWh and Sense recording 83.3kWh.
The differences seemed rather suspiciously consistent. Not sure what that means.
Thanks for all your help in helping me get to the bottom of this conundrum.
Hayward
This is one of the nicer, neater panel wiring jobs that I have seen posted in a long time! Someone took extra pride in doing the job correctly.
The only comment I can suggest is to tape the CT clamps closed using masking tape or painter’s tape. When you install the metal panel cover back on, it might be pressing against the clamp release button and causing the clamps to slightly open. Your clamps appear to be centered on the incoming service wires and there doesn’t appear to be any interference of other wires that might create a problem.
OK, I’m going to go back to my earlier graph to help you analyze your results. I’m hoping that you will try to do the same kind of graph in Excel (or Google Data Studio), a scatter chart of hourly Utility vs. Sense data. If you look at chart I did, using the data you posted you will see three things.
The data - in points on the graph
A blue fit line based on linear regression, a way of fitting data to a line
A bluish band around the line - that’s a 95% confidence interval (ignore that for now)
The blue fit line is a line that fits your Utility vs. Sense data best (smallest “error”). In an ideal world, with Sense matching your Utility, your results (OK with solar) will look like this, a 45 degree unity line with almost all the points on that line.
Your slope is about 0.83 instead of 1.0. That means for every 1 kWh your utility usage increases, you only see a 0.83kWh increase in Sense’s number. You’re seeing when you have 98kWh of Utility power showing up as 83kWh in Sense for a day.
Your line does not go through Utility = 0, Sense = 0.
Your data points are scattered on either side of the fit line in sort of a random pattern. Part of that might be due to the fact that your meter doesn’t seem to give you an additional decimal place of accuracy for your hourly measurements, while Sense does.
Issues 2 and 3 are more problematic than 1, in my mind. If you were getting a straight line with a slope of 0.83 and all the points were very close to the fit line, then something systemic, but likely fairly simple, would be wrong. But having the points all over the place means that you are getting random variation - much harder to track down.
I’m willing to help analyze your data as you collect more of it. And you can send/post csv spreadsheets via this forum.
I still would suggest that if your meter displays current power usage as part of it’s cycle, that you take your phone out with the Power Meter on and watch / compare the power number. If the Power Meter shows power that is 0.83 of the usage showing on your utility meter, than one of the two is miscalibrated or behaving badly for some other reason.
Thank you so much for your suggestions and responses. I believe I have found the source of the problem. I wrote the ‘Sense’ support people and they suggested the sensors were not closed properly or may have become nudged apart when I put the cover on the electrical panel. So I took the panel off for the following few days and as a added assurance I taped the sensors together as suggested by this community.
I’m happy to say that over the 72 hour period I did the meter readings and compared them to the detail Sense data there was less than a 0.1% discrepancy.
Thanks to everyone who offered their insights and help.
Hayward
This!!! I had the exact same problem! Pulled off the cover and there was indeed a small gap on the sensor. Removed it and put it back on. Taped it. Put the cover. Dropped again. Pulled the cover and I could see where the tape pulled apart. Repositioned stuff and now the Sense matches the meter! Awesome!
