Experimenting with predai

[SwiftRR]

Hi Trefor,
I have experimented with predai on my raspberry pi HAOS setup. I have free-standing predbat installed without appdaemon.
I have got it working and have added the relevant lines to my predbat apps.yaml.
I managed to generate an apex chart that looks very similar to the chart shown in the documentation.

I did notice that there was an extra space in the documentation code lines that I added to my apps.yaml file. It reads as below.
I removed the extra space.

load_forecast_only: True
load_forecast:
- sensor.givtcp_{geserial}_load_energy_today_kwh_prediction$results

Some questions:

1. I wasn't sure what to add for the 'subtract' line for ev use which refers to a 'sensor'. In predbat, ignoring ev charging or large usages is handled by a switch or inputs (in documentation car charging hold options). I have a zappi.
2. Should I change the historical days to 10 in the predai yaml to match what I have in HA currently? In fact, what use is such a small sample of historical days? I see that you want a larger sample in a database.
3. I have a heat pump with weather compensation, and I have been able to add the sensor for my weather compensation to the predai.yaml.

This is what happened to my load predictions for Monday 12th August – scary with so many predicted loads of zero.
My predbat settings are nothing out of the ordinary.

I've now commented out the predai code lines from my predbat apps.yaml.
I have kept predai running though. Should this then start populating a historical database?

Still an interesting experiment and good to see that is seems to be quite easy to step back.

Rob

[SwiftRR]

Trefor,
as an add-on to my earlier comments in #13, I have checked predai.db and I get this error:
'utf-8' codec can't decode byte 0xf8 in position 102: invalid start byte

So it looks as if a database isn't being generated (at least for me). In your video, you suggested that a database would be for the future, relying on HA for now.
This is my predai.log file:
predai.log.txt

In the predai add-on log, I get lots of warnings referring to neuralprophet and types used, e.g.

WARNING - (py.warnings._showwarnmsg) - /usr/local/lib/python3.10/dist-packages/neuralprophet/df_utils.py:1152: FutureWarning: Series.view is deprecated and will be removed in a future version. Use ``astype`` as an alternative to change the dtype.
  converted_ds = pd.to_datetime(ds_col, utc=True).view(dtype=np.int64)

For completeness, these are the apex charts that I have generated for load and temperature:

I intend to leave predai running as an add-on and serving predbat for load.
I will see how it proceeds.

I hope my comments are useful.

Rob