long term statistics for binary sensors.

[eliasel]

Describe your core improvement

To have the long term statistics database also save data for binary sensors.
My suggestion is to save the to aggregated datapoints for every hour:

• "The value at the start of the aggregated time" (hourly).
• "Average" of a binary sensor (i e. considering "off" = 0 and "on" = 1) and computing an average sensor for each hour.
• "number of transitions" the number of times the sensor has transitioned from "off" to "on"

With the suggested setup above I believe it would be possible to re

Current limitations

The curreny limitation is that as far as I understand currently there is no long term statistical data saved for binary sensor. Which is alot of data being lost. As an example I have a heatpump which I log the temperature going into the hot water accumulator. I have a binary sensor for the trend of the temperature, but that is not saved beyond the standard 10 days.

Technical benefits

Possibility to use binary sensors for data analysis.

Possiblity to do daily/weekly heatmaps of when the binary entity transitions/ is active.

Additional context

My initial reason for this request was that I would like to understand at what times during the day the heat pump is active. ( I also have the exact same usecase with a my shower being on).
For both situations I have a binary sensor which logs when the element is active, but since it is binary sensor there is no data saved beyond 10 days. This for me is a technical oversight since the binary data already is a very compressed data format.
But I understand there may be binary applications where there might be enormous amounts of transitions and such cases may risk filling up the statistics database.