Add simulated power sensor (along with the simulated energy sensor)

[cybermaus]

Nice utility. Does pretty much exactly what I was looking for.
This enhancement suggestion is not important, just a nice to have, it only adds ability for power graphs

If I want a power graph, I currently have to calculate simulated power, and I did so by subtracting charge/discharge power from grid power, with a zero floor. And this works, see below, except because both of these source devices have their own intervals and sample moments, there are fringe effects which cause spikes on the edges.

My guess is if the simulated power was calculated internally, the fringe effects would be only half as strong, because calculations would match at least one of the devices for interval and sample moments.

And BTW, I do understand that the actual simulation is using energy, so does not suffer from these fringe effects, because it is using and resulting in cumulative (total_increasing) energy. So the core of the battery_sim results are fine. (Almost too perfect, considering a real battery system also would have response times)

Here is the power graph I made with my self calculated power, showing the spikes caused by the fringe effects.

Edit: I see this is similar to #136 except I think I am explaining/asking in a more abbreviated manner. Not sure if that is good or bad, but it is a 2nd viewpoint.

[dewi-ny-je]

This simulatior works by analysing energy flows, not power values (at least for now). Even if the simulated power flows are provided by the integration, they would not be more accurate than the power flows you can get from a helper sensor.
In fact, since this integration updates values by default (and by strong recommendation) once a minute, you would likely get more dynamically updated graphs by using said template.
Your smooth graphs with short spikes would become staggered graphs with 1 minute accuracy...

So I'm not convinced this change really makes much sense.

[cybermaus]

Edit: I got rid of most of the spikes by setting both my utility meter (P1) and solar inverter (SolarEdge) from 30 seconds to 3 seconds. Strictly speaking they still give interference spikes, but they are now avg 1.5 seconds long rather then 15 seconds long

It does cost 10x more data of course. I tried 1 and 2 seconds, but it started getting irregular on the read-out.

[dewi-ny-je]

To see if the spikes are really caused by the issue I reported, you could also try setting the DSMR to 0 and then add a filter which first averages and then throttles:

sensor:
  - platform: filter
    name: "Throttled Average Sensor"
    entity_id: sensor.your_original_sensor
    filters:
      - filter: time_simple_moving_average
        window_size: "00:05" # Window size for average (5 minutes here)
      - filter: time_throttle
        window_size: "00:05" # Output the result every 5 minutes

You have to add it to configuration.yaml, manually.

It's actually a good test case for the bug I was trying to correct. If it works I'll use it to strengthen my point for my DSMR patch

[cybermaus]

Ok, just tried setting 0 from your previous not, I did not know that was possible, but that at least works.
Will do the yaml later. I actually already have a wrap around sensor, so it should be is easy to add.

[cybermaus]

Just some intermediate feedback: did not do the filters yet, wanted to watch the result of the 0 yet. And so far it seems the spikes are pretty much gone. I still see some, but they are actual power user spikes from my Quooker, and not phantom spikes due to interference/interleaving on the readout moments.

Mind you, those Quooker spikes, strictly speaking they should come out of the battery, but I think there is a 60 second value on the battery_sim as well?

Edit:I have a 3-phase system, that is why the Quooker does not fully come from the solar, it can only use 1/3 of the solar. I need to install a balancer from SolarEdge

[dewi-ny-je]

The battery gets updated every 60 seconds but since it operates on "total energy flows since the last update" and not on instantaneous power, it will take care of spikes perfectly.
In fact, better than a real battery which does averaging and takes 1-3 seconds to react to sudden power changes. If you check battery reviews they actually test the Quooker and its use causes a worse "zero at the meter" behaviour.

About 3 phases: don't spend money, the energy company should only look at the sum of all phases so they don't care if you pull from one and push to a different one, they care about the net flow.

[cybermaus]

Re spikes, yes, I understand. Energy wise, the solution is quite correct because it works on bookkeeping of kWh.
The spikes are only on power, and then even on artificially derived power, because I want some nice looking power graphs. But they are mostly aesthetic. Which was my original request, if the power is derived internally, it is one less source of artefacts due to out of step time windows.

As far the 3 phases question goes: I know, I did quite some digging a few years back, to the point that i even read the specs and know one utility-meter brand required a firmware upgrade to get into spec.

However, I do not trust 'them'. Especially when I see stuff like this:

Ok, this is again power, not energy. I remember back a few years back, I check the output of the kWh for a couple of days to see if there ever was a single second where both in and out increased, and it did not. So all good. But they can change the rules. It may need some lobbying, but already they managed to lobby the "saldering" away.