Multiple total counters have negative values regularly
TimelessNL opened this issue · 7 comments
Hi,
First I want to say great work so far!
I was working on a GoodWeLogger setup but then I came across this project.
Since I already had the WiFI dongle I figured this project/library based on UDP was worth a shot.
I've connected HA to my GW5K-DT and everything seems to work thus far.
Fortunately my e_day and e_total do work kinda, but they report -0.1kWh regularly:
This is also the case for the h_total sensor, seems like a problem with sensors reporting total values?
So my question is, is this a known issue yet?
I've enabled debug logging, And during a -0.1kWh moment there are no errors logged.
Note:
This question was first posted in #71 but I moved it since the protocol logic is part of this library.
Well, not really sure how to cope with this.
If it is reporting non-zero (not None) values, it is definitely not the library code, so it must be the inverter which provides these odd values.
(Or in theory the library may decode wrong packet, but in such case all values reported at that point in time would be off usual values. Can you observe such correlation between various sensors ? Or is it just e_total which is off ?)
Can you observe such correlation between various sensors ? Or is it just e_total which is off
As for now e_total, e_day and h_total are all negative at the same time (maybe more).
I have not verified ppv and temperature with the inverter, but they look plausible during such an event.
For example this was today's temperature reported by the inverter:
And this was today's e_day reported by the inverter:
As you see e_day repeatedly goes to -0.1kWh shown by the blue filled graph, while temperature has a steady line.
I can't think of a reason why the inverter would send negative values for these kind of parameters.
My inverter is not connected with the internet (firewalled) but it would be strange if that's the fault.
it is definitely not the library code, so it must be the inverter which provides these odd values.
Maybe a workaround or sanity check could be added?
For example when these parameters are reported negative return None instead?
This way homeassistant can skip the sensor update, and will get its update at the next interval.
I haven't seen it on GoodWe but other Modbus devices, that they return -1 for "no value". the e_total for example is read as 4 byte signed, therefore the value on the wire would be 0xFFFFFFFF. Maybe this could be detected and as @TimelessNL suggests a none could be returned.
Yes, that sounds reasonable, at least for the 4 bytes power/energy kinds of sensors.
I'll adjust the library code to ignore such -1 values
What about the h_total sensor? That one seems to have no value results as well.
Had some issues with my utilities (they remote-shutdown my inverter, which then went to "Off-Grid" mode. Took me a while to find the root cause.
During the process I restarted the inverter a few times and during two restarts, the following total values were read as "0", which screwed up the statistics for the "Energy" Panel.
Affected totals:
sensor.goodwe_e_totalsensor.goodwe_e_bat_discharge_totalsensor.goodwe_e_bat_charge_totalsensor.goodwe_meter_e_total_impsensor.goodwe_meter_e_total_expsensor.goodwe_e_total
As these values are ever increasing, would it be possible to ignore "0" readings?
-1 handling value was implemented, filtering 0 values for total_increasing sensors was done in HA code.