Solar customers: Recognizing the "Daytime dip" and how it might impact your Sense numbers

We’ve (well, I’ve mainly, with other people rolling their eyes more than likely) had some discussions on solar production calculations and errors therein and I just wanted to consolidate it to one thread that might hopefully help solar customers recognize the issue and determine whether or not it’s affecting them. According to @HilarioAtSense in this post, Sense has recognized the problem and are working on a fix, pending identification of the users affected by the issue.

This thread is meant to be a summary of my findings and the means by which other solar users might identify whether they are affected or not, starting with a very quick and simple method and progressing through a couple more labor intensive methods. I figured that providing the data and how the calculations were made would allow other solar users to better grasp what I’ve been babbling about for the past two months.

The least labor intensive way to identify whether you’re affected by this is to take a look at your usage graphs from a distance, by manipulating the graphs to cover 12+ hours. What you may notice, if you have a home that’s typically quiet at night when you’re sleeping, and during the day when you’re at work, is that the baseline power consumption is different, and notably during the day takes a bit of a dip that’s correlated to the magnitude of your production values. A composite of the past 48hrs from my Sense unit is shown below. Note the very consistent baseline during the night - that’s my fridge mostly humming along, in contrast to the “smile,” or “daytime dip” produced during the day, where that baseline drops in tandem with the rise in solar production. This is because the solar CTs are improperly calculating the power coming in from the solar panels; the power being calculated from the main CTs is spot on, so while it’s accurately measuring the power going back out through the main CTs to the grid, Sense is secondarily miscalculating the power the house is using (because the information it’s getting from the solar CTs is incorrect).

That’s the easy way to identify the issue, but it requires a pretty calm house at night and during the day to identify. Another way that’s a little more labor intensive and requires a secondary measurement device - possibly your inverter if it’s able to provide the information, is to compare the solar production graphs between the Sense device and your secondary device. This was actually my first indication there was an issue. I went back through my Sense readings and identified the production values on a 15min schedule…mainly because that was as frequently as my inverter was reporting values. I then graphed my inverter 15min readings against my Sense 15 min readings, and over the course of a few days what became obvious was that Sense was reporting production values about 3.5-4% below what my inverter was reporting:

The most labor intensive way I was able to ferret out the problem was by taking twice daily readings of my meters, just before sunrise (on the hour!) and just after sunset (on the hour!), to make sure I was separating out the numbers achieved when the sun was up and we were making most of our power vs. when the sun was down and all of our power was coming from our utility. This requires a bit more effort, and a meter (or meters) which allow you to determine (1) your solar production, (2) how much of that solar was sent back to your utility, and (3) how much power you are using from the grid. In my case, I have two meters - one that measures solar power and a number of other solar-related values, and a second that measures how much power we’re sending back to the grid and how much we’re using from the grid, among other grid-related values. You will have to check with your utility to identify how you may do this at home using your meter(s).

By taking pre-sunrise and post-sunset readings of my meters (on the hour, because that’s how Sense divides things!), I was able to build a spreadsheet (linked below) to develop calculations that allowed me to chase down where the error was occurring. You can use this spreadsheet as well, if you have access to the data I mentioned above. You will have to enter your own meter readings, and laboriously calculate the Sense values from your daily trends graphs by pressing on each bar and getting it to read out a value for you for that particular hour.

To fill out the spreadsheet, you’ll enter your solar production meter info, your utility power reading, and your solar out reading. Your solar production, energy in, energy out, and total energy usage will be calculated from these meter values automatically. Next, you’ll enter the data you calculated from your Sense trends. Enter your Sense solar production values and your Sense energy usage values for the daylight and nighttime periods. The rest will be calculated for you - the discrepancy in production (in my case ranging from ~3-4+%) and the raw values. What I found was that, again, Sense was under reporting how much solar I was producing, and that was directly impacting how much Sense was reporting as used during the day - the raw values were equivalent, meaning that if Sense was low by 1.5kwh on solar, my daily usage was also low by 1.5kwh. (See the values for “usage discrepancy raw error” and “Solar Sense raw error” highlighted in pink)

A link to a Google spreadsheet is here, because I’m not able to upload a spreadsheet. You should be able to copy and paste it to your own xcel file, but please let me know if you’re interested in doing so and cannot - it only has permission to view, not to edit, though that should still allow you to copy-paste. The information you’ll need to enter on your end, apart from the date, is highlighted in red.

If you’ve gone through the first two steps and found obvious differences between what Sense is reporting and what your inverter or meters are telling you, the third step should help you confirm what you’re seeing. You’ll be looking to see that nighttime values are matching up well between your meters and Sense, while daytime values are off. You should see that the raw value for solar production error is equivalent to the raw value for daily usage error, highlighted in pink on the spreadsheet.

Thanks for following along, and let me know if you have any questions or comments.

2 Likes

For those who may be wondering why this might be relevant to them: I’ve spent some time, now that I have a couple days off, confirming this does seem to affect device detection. A couple appliances and a set of lights that are recognized before/after daylight didn’t seem to show up when we’re generating power during the day.

There’s some beta feedback Sense should be taking seriously - solar customers will surely be at a disadvantage when it comes to consistent appliance detection.

1 Like

[quote=“NJHaley, post:2, topic:403”]
There’s some beta feedback Sense should be taking seriously - solar customers will surely be at a disadvantage when it comes to consistent appliance detection.[/quote]

I tried an experiment yesterday, and it does seem like my Sense monitor is underreporting the solar production by about 2.5%. Now, 2.5% may not seem like a lot, but when you’re subtracting two large numbers to get a small number, a small percentage error in the large numbers can translate to a large percentage error in the result.

Yesterday my solar panels were generating about 7kW, with a small amount used in the house, and the remainder going to PG&E. Sense reported that the difference between the solar production and the power going to PG&E (i.e., the power used by the house) was 317W. I turned off the solar inverters, and the power used by the house as reported by Sense jumped up by 180W to 497W. That’s a 36% error in the reported power used by the house.

1 Like

Thanks for verifying, for a while I thought I was going crazy! Thanks for posting another relatively simple way to ID whether you’re affected.

And you’re right - the percentage may seem small but the raw values will add up to about 50kwh/mo in the summer time for me.

I don’t know about you, but except for perfectly clear days I see quite a bit of variation in my solar output as clouds and haze impact it. Thus, I don’t think I would be confident of the baseline with solar versus without it.

[quote=“NJHaley, post:4, topic:403”]
And you’re right - the percentage may seem small but the raw values will add up to about 50kwh/mo in the summer time for me.[/quote]

You are right, the absolute error is an issue, but I’m more concerned about how this affects device detection.

By identifying some of the power hogs, I hope to get my always-on load down to 250W.

Suppose in the summer the solar panels are producing 10kW. The always-on load is taking 250W, and the balance of 9.75kW is going to PG&E.

But if the Sense monitor is reading the solar production 2.5% low, it will be registering 9.75kW solar production, 9.75kW going to PG&E, and therefore zero consumption in the home.

I think the Sense monitor algorithms will have trouble reliably detecting loads if it’s reading zero when actually 250W is being used.

Thinking further, to 11kW production and 200W always-on, the error in the sense reading will be greater than the entire load it’s trying to measure. That can’t be good.

Ding ding ding! In fact, your always on turns to zero, which is something I started noticing as our days started getting longer and less cloudy:

Prior to a couple weeks ago, we were pretty cloudy since right around Xmas (when I started up my replacement Sense device), not making too much more than we were using. Back then I posted a thread winning “always on” value: So what is your MINIMUM power usage? 249 here for a big house - #2 by NJHaley Now I’m at zero :slight_smile:

I think this horse has been beaten - now to hope Sense makes strides to fix it.

Here are a couple of other examples of the daytime dip from my system.

In the first example of a daytime quiet time with no one home you can see I have a very stable usage pattern. The main usage is from a variable speed pool pump from which you can see a very clear tooth pattern normally. The tooth pattern is mainly from the variable speed motor switching speeds to adjust for the pool solar water heater. The interesting thing is to see the highly irregular pattern occur along with big solar fluctuations from passing clouds.

In the second example you see a more pronounced dip occur around 2:30 in what would normally be a very stable time period.

1 Like

Interestingly, yours seems to be more of a “blip” than a “dip.” While I would expect mine to rise when the sun goes behind a cloud, yours drops when production goes down.

Why it’s the opposite of mine is escaping me - do you have supply-side or load-side solar? Mine is load-side for what it’s worth.

I’ll take a guess that the difference of a blip versus a dip “might” have something to do if they had to do a software flip to account for a CT that was in the wrong direction. I was just looking at my system today based on what you are seeing and I am most definitely seeing consumption go UP when the solar goes DOWN due to a cloud. This is not good and they need to figure out what is going on here ASAP!!!

Here are a few pictures:

I’m curious about the software flip too, @Howard. As best I can tell, mine had to be “flipped” since my CT decals are facing my inverter, not my main panel.

For posterity, though, are you load or supply side solar? (ie is your solar feeding into the panel through a breaker, or is it going directly into the mains?)

My solar feeds through a breaker in my main electrical panel.

I tried the experiment of turning everything off in the house this afternoon.

The Sense monitor should therefore have been reading 0W. It actually showed negative power use: -78W.

The Sense monitor reported that the solar panels were producing 3146W. They were evidently producing 78W more than this at the time — about a 2.5% error.

2 Likes

As a new user, I am experiencing a problem that may be related, but not exactly the same. After 3 weeks and several problem reports my tracking solar arrays seem to be reporting OK but now my Mains power is way too high. The Mains readings seem to be 2X the solar power minus the actual power used off the Mains.
Are the Sense technical team looking at each “new” installation and tweeting the power reporting procesd?
Quick edit: Just took another look and one Mains leg is reporting almost all of the power from the two solar legs. Strange!

1 Like

I had a similar issue recently happen with my system where usage was about two times production minus reported consumption. They have to do a total reset of my system from scratch and wipe everything out. :frowning:

Wierd, I have solar, looking at my logs. I do not have this problem.

Do you have any details on your install? Sounds like you’ve got an enphase inverter - is it tied in as supply side or load side? Any specifics on how your cts are arranged?

Early adopters get snubbed again I guess :confounded:

If it’s something as simple as how it was installed or the Sense build, I wish they’d just say so. I’ve done just about everything I can to correct this, including a new unit; it’s ridiculous it hasn’t been fixed yet. This is honestly the most terribly supported smart device I have.

I do have enphase inverters, and it is tied in with the loadside. I’m
wondering if maybe the inverter is too close to the sense CT clamps.
Eletromagnetic interference maybe? Have an RF meter? I was an early
adopter as well, so I don’t think that’s the problem. I’ve had a few issues
but the sense team works them out fairly quickly, or fixes it before I
report them. Wondering if sense will need to create some sort of
calibration setting when this scenario presents itself.

I feel, based on some of the Sense team posts, that it’s a legitimate software-side problem related to a voltage calibration problem they pushed a fix for back around Xmas. I think the trouble they’re having is identifying who’s affected, based on their comments.

I know that one of their competitors has a setting that the end user needs to set dependent on what sort of inverter connection the user has, but I have a feeling this issue is a little outside of that. My nighttime measurements are dead on, it’s only when solar is involved that things get mucked up… unfortunately as the days grow longer, that has more and more influence on the reporting as a whole.