FoxESS Community - Owners & Installers Forum

Hi

I have a KH7 inverter with an EP11H 10kWh battery and 10 Aiko 4.7 kW panels. As far as I can tell, it works just fine and generates plenty of electricity on sunny days and less, but still useful amounts, on cloudy days.

However, since the beginning of March, something seems to have changed in the way the data are reported either on the FoxESS website or via the API.

The first thing I noticed was that the time at which PV generation started suddenly got later in the mornings, at least as reported on the website. In the last few days of February, generation was consistently starting before 0800, and closer to 0700 on sunny days. In early March, generation didn't generally start until about 0830 and sometimes not until after 0900. For context, sunrise was at 0649 where I am.

I looked into this a bit further and found something strange was going on.

If I want to know what my daily PV output is, I can get this information in one of 2 ways. If I go to the FoxESS V2 website I can download a CSV file of daily PV output.

Alternatively, I can download data from the API, which gives me the instantaneous PV generation every 5 minutes or so. If I calculate the average of those readings over the course of an hour, that should give me an estimate of the total amount of electricity generated in that hour (average generation in kW for 1 h is the same as the amount of electricity generated in that hour in kWh), and then I can add up over each hour of the day to give me the total generation for the day.

Now, that second method won't be perfect, as it's relying on readings taken at a point in time which won't necessarily reflect the true average over the hour, because PV generation is fluctuating minute by minute. But any discrepancies are just as likely to be in one direction as in the other, and averaged over the course of the day you would expect such discrepancies to largely cancel out, and the total figure should be a pretty good, if not necessarily perfect, estimate.

Up until the end of February, those 2 methods of calculating the daily output were in almost perfect agreement. However, starting from the beginning of March, they really weren't. When I calculated the ratio of the daily output as calculated from the API to the daily output as downloaded from the V2 website, it was 99% in February (and always between 95% and 104%) but only 65% in March (ranging from 19% to 89%).

I think the problem is shown pretty clearly in the attached graph, in which the 2 sources of daily output data are plotted against each other by each month since last August. You can see that the values agree very closely indeed in every month except March (apart from just 1 day in December when something weird seems to have happened, but I'm not going to worry about a single day), and March is clearly different.

I have opened a ticket with FoxESS support who completely refuse to believe that anything is wrong here, and have just said that my method of calculating the daily output from the API data isn't valid. Well, clearly it isn't valid any more, but it certainly was valid up until the end of February.

One other piece of information that may be relevant is that I set my battery to charge overnight using the cheap electricity rate, and this normally works just fine, but it didn't charge in the early hours of 1 March, which makes me wonder if there was some kind of software update at that time which may have broken something.

Does anyone have any idea what's going on?

Thanks
Adam

I don't know the answer but Fox regularly change the way statistics calculate, here it looks to me like the 5 minute samples you are averaging may now be taking system losses into account in their readings.

I have an old H1 hybrid and one of the 'benefits' is that Fox never release new firmware which messes with the statistics - when I look at my HA dev system for Fox OpenAPI, my PV yesterday is calculated using the 5 minute averages into a Rieman sum and I see 14.2kWh, whereas the Fox cloud reports 14.4kWh which is relatively close.

As you are using the OpenAPI to gather data and you have a hybrid inverter, as a confirmatory check when you get the 5 minute sample from the endpoint '/op/v1/device/real/query', you can look at the variable 'PVEnergyTotal' - this is what the inverter itself is reporting as the PV production for the current day.

Thanks for the suggestion. On my system, PVEnergyTotal gives me the cumulative total energy generated since it started counting rather than a figure for the current day, but it's easy enough to calculate the daily PV production from that by just subtracting the value at the beginning of the day from the value at the end of the day.

If I do that, then I get exactly the same numbers as the ones I download from the FoxESS V2 website.

So it may well have changed the way the statistics calculate, but it looks like it's only changed it for the instantaneous values, and not for the total values.

Very strange.

As per viewtopic.php?p=13287#p13287


They changed it in a recent firmware update. The power sensors now have about 80-100w subtracted, which hides the parasitic loss from the inverter/bms/etc, but the solar generation sensor will show the original. Previously the 80-100W was added to the household load sensor. So sometimes you will see the solar generation sensor increasing while the power is still reading zero.

Similarly, you will see that the PV sensors are nonzero for voltage and amps, but the power sensor is showing zero. Also, if you set up a riemann integral for the power sensor, it will significantly under-read the solar generation today sensor (the volts, amps, and solar generation sensors are consistent with each other, but the power sensor is not consistent with them).

If you have home assistant, you can set up a template sensor on sum(volts*amps) for all the strings combined if you want a power sensor with the old behaviour.

TBH I wish they would just create a losses sensor instead of trying to hide it in various other sensors.

Going back to the load sensor, the situation with the solar sensors now, is similar to the situation with the load sensors on the old firmware (the load energy would read high compared to a riemann on the load power sensor, because the parasitic losses were being added to load energy). After the update, the riemann on load power is now very similar to the main load energy sensor.

Thank you! That explanation fits perfectly with all the observations, so I'm sure that must be the correct answer.

I checked the volts and amps, and before the change, they matched up exactly with the reported power. Now they don't. Though the difference is more than the 80-100 W that you mention: I have 2 circuits (PV1 and PV2), and the difference is more like 150 W per circuit, so 300 W in total.

Not sure if that's just normal variation from one system to another or if I need to get my installer to check why my inverter is so power hungry?

I also noticed another change that happened at the same time, which all of a sudden makes sense. Previously, if I looked at the solar generation, flow to or from the battery, flow to or from the grid, and the house load, they didn't quite add up. I assumed that was something to do with the power the inverter itself was taking. Since the change, the figures add up perfectly. If the losses to the inverter are now taken into account in the reported power generation (and I agree it's not the most intuitive way to do it) then that completely explains why that behaviour changed.

Weird that FoxESS technical support didn't seem to know about any of this. You would have thought they'd know what their own firmware updates do, but there you go.

I too have seen an INCREASE in my standing load...prior to the install just last week my smart meter showed it was always around 200 - 250W, now, on the app and webpage its nearer 300 - 500 even overnight when pretty much everything in the house is off...cant really see why though, even when the battery was below 10% and therefore effectively off the smart meter showed the house standing load was over 400W which is not right