Inconsistent average

The day view showed an average of 864W in the evening for my solar.
The following morning, the average was 713W.
Given that no power was generated overnight, and the only difference between the rrdtool output was a translation of the solar graph, why is the average different?

I was expecting day view average * 0.024 would be close to the daily energy reported by the inverter, but instead the flukso average is both high and inconsistent.

icarus75's picture

Hi Ghostgum,

The day-view average is a rolling average of the power consumed or produced between now and now-24h. So there might indeed be a difference because in the morning yesterday's production would already have partly fallen off the graph (and replaced by today's).

Day view average (W) / 0.024 = kWh/day


ghostgum's picture

The problem is due to the solar dropping to zero overnight, but flukso not recording this in the graph. The average then climbs overnight, from 918 just after sunset, to 1836 the following morning. I'll email you some graphs.

ghostgum's picture

It looks like the flukso sensor stops sending data to the flukso server when the reading is zero, as it is for solar overnight. rrdtool appears calculate the average from the available data, which is less than 24 hours for any time between sunset and sunrise. Average * 0.024 is not the daily energy for these times.

How do you convince the flukso sensor to send the readings of zero energy to the server?

icarus75's picture

Regardless of using the clamp or pulse input, we work with 'Wh' energy pulses as a basic measurement unit. The daemon on the Fluksometer adds these to a counter and sends the (timestamp, counter) pairs to the server. The (timestamp, counter) pairs are then entered into an rrd with its data source type configured as DERIVE. When your inverter stops outputting power from the PV panels, no current will be detected and hence no Wh-pulses generated, no counter updates and no entries into rrd. This is the reason why you will not see an update until sunrise.

One option would be to generate an artificial (timestamp, counter) pair with an unchanged counter value when not receiving a Wh-pulse within a certain time interval. This would trigger a 0 Watt line on the graph. But what if a fraction of a second after this event a Wh-pulse is generated? It would show up as a huge power level in the graph that cannot be accounted for. So I don't think this is an option.

For a correct calculation, you could use the API and calculate the watt average from there. As for the rrdtool average calculation, you can't really blame it for only using the available data.


ghostgum's picture


If an artificial Wh pulse was generated, then you would need to remember that and subtract 1Wh from the following pulse, and hope that the following pulse wasn't a multiple Wh pulse.

I was trying to compare the Flukso measurement with that reported by the inverter. Flukso seems to report about 10-15% higher than the inverter.

somm15's picture

Day view average (W) * 0.024 = kWh/day

Rather than
Day view average (W) / 0.024 = kWh/day

Isn't it?
It xW during 24Hour divided by 1000 no?

somm15's picture

I think I could have a related problem.

I realised that the power measured jumped from 100W to 250/300W when I powered my heating plates. Well, apparently it's only the power measured by the FluksoMeter not the real one.

If you have a look at my graph, you'll noticed that according to the Flukso my minimum power during last night was 260W.
I took the numbers from my electrabel energy counter yesterday at 11PM : night = 923.1, day = 925.3
This morning at 7:35 AM the numbers were: day = 923.2, night = 826.1
So, normally during the night the power consumption was 0.9KWh. Considering that there was 8.5 hours between the measurement, the average power should be 106W.
But the Flukso says 280W (more or less). Strange.

Then, I reviewed the excel sheet with the numbers from the electrabel counter I take each month. My average consumption since september 2010 is 6KWh.
If the eating plates are really eating 200W continuously, it means 4.8KWh each day. It means only 1.2KWh for the rest. It means 50W continuously and no other consumtpion.
Finally, on my monthly graph (somm15) you can see a day without the eating plates, the average is 240W. It means 5.8KWh.
There is also a day with the eating plates, the average then is 440W or 330W. It means between 10KWh and 8Kwh. It's far more than my actual consumption.

All this boring explanationg to say that I think that there is a problem with the current clamp and the heating plates.
According to a forum and a friend, this could be because the power supply is far to complex and they explained things that I don't understand with 100Hz Harmonics...
The conclusion is that it's impossible to measure my overall consumption because of the heating plates.

Another funny thing to confirm is that the heating plates consumption seems to vary if I switch on and off the power (160 - 250W). I can have 200W during when day and then if I switch off and they on the power, I'll get 250W measured the next day.

I'm a computer scientist not an electrician but I think it could means that the 100Hz complex explanation could be true.

If you have any question, don't hesitate.

I'll do another comparison with the electrabel counter tonight.

skynetbbs's picture

are your cooking plates connected on the 400V power supply or the 220V?
are they using "induction"?

the clamps only monitor "current" and do not measure "voltage" will allways think it's 220V ... not even notice variations between 210-245V

Also it doesn't see the direction of the "current"... have you installed solar panels?

there are known issues with the clamp when using TL lamps
(something about cosinus phi)

do read this discussion:

a solution would be to forego the "clamps" and use an expensive measuring tool and use it's digital output connected to your flukso

somm15's picture

Thanks for the response.

It's a normal 240V power supply, no solar pannel.

I use induction heating plates connected to a 32A 240V circuit. The current clamps is connected the before the main circuit breaker.

I think I'll try to use one of these devices to measure my overall consumption:

Do you think it'll work with the port 4 or 5 of the flukso?

Apparently, Whirlpool heating plates do not work with the normal flukso current clamp...

somm15's picture

If got a response from Bart.
Both devices should work but he would go for the siemens one because it has two S0 pulse outputs: one for real power and one for apparent power. I'll by one of these next month and post my feedback.
He also suggested cheaper chinese devices:

icarus75's picture

The Siemens one is indeed nice as it outputs real and apparent power via two discrete S0 ports. If however you're more price- than feature sensitive, you might consider the ones in the above link.