WTH is the diffrence after last state change on my temperature sensors?

Is it posible to get “vektor” attributs on all my numeric sensors?
e.g for my temperatursensors, what is the diffrence after last state change ?

  • Is the temperature state (the Amplitude) rising or is it falling ?
  • How fast is the temperature state changing (The Frequency).
  • What is the temperature diffrence (length/distence) between last state change (The Velocity)

Is it just me, or would this be useful when you have a numeric sensor like power consumption, humidity…?

If I am not mistaken, this should be possible to calculate with simple algebra functions, Sine, Cosine and Tangent from the state change between the “Amplitude” values.

It shuld also be posible to predict “guess” the next state change value, of the temperature.


Math inspirational videos on the subject:

It’s an interesting Feature Request but it means all of those calculations would be performed for each sensor every time its value changed. That’s a significant amount of extra processing effort and, at least in my case, I don’t need all of that data for each sensor.

Perhaps it could be offered as a per sensor basis. If I need the extra information for only five out of 50 sensors, I can enable an option for each of the five sensors.

1 Like

As @123 stated, applying those calculations to every sensor could have a high processing cost for users with more than a few sensors. Most, if not all, of what you want is already available through the various built-in sensor platforms. Take a look at the Integrations page especially the Utility group where you will find a number of statistical and mathematical evaluation sensor platforms.

As @Didgeridrew suggested, derivative has a temp change example and the statistic sensor attributes offer change average change and change rate

I do use them today, but hope there can be an easier and better solution in the future:)
At least to get an attribute that say if the data it is trending up or going down.

Adding one of the mentioned sensor integrations isn’t difficult and constrains the additional processing to selected sensors (as opposed to all sensors).

In addition, if you look at the Trend integration for example, there are several options available to influence its behavior. It means all of those options would also need to be available to each and every sensor. The amount of configuration per sensor wouldn’t be less than the current way it’s done.

I agree, most of it is awailable today with templates and the Utility sensors, espesially if you are a programmer.

It would be nice if there was an easyer way to implement more machine learning logic into the automations for non programmers also.

Perhaps there is a possibility to one day get an option to generate related virtuall sensors, based on the platform’s that consumes the state of the sensor.

Filling out a few options isn’t programming. Anyone who thinks it is programming should reconsider their choice to use Home Assistant because there other aspects of it that are likely to overwhelm them.