Aggregate string time series

Hi @jtfrandsen!

Thank you for suggesting this product idea. We will follow the traction this idea gets in the community. You can expect updates on this post if we decide to include this in our future roadmap, or if we require more information.

Hi @jtfrandsen!

I’m the product manager for our Time Series service and your question is very timely, as we are planning to introduce a new “first class” type of Time Series to provide better support for machine state data.

The aggregations you have described here are exactly what we plan to support.  In addition, we will support the definition of machine types (proper name to be confirmed) in which the state names will be defined for given machine types (as opposed to the use of arbitrary strings for state names).  We plan to have well known machine types defined out of the box (i.e. OPEN/CLOSED, OFF/ON/STANDBY) and provide customers with the ability to define their own machine types in addition.

We expect to commence development of this feature in late Q3 / early Q4 with a release in Q1 2025

In the meantime, might I recommend the use of the numeric type to represent states.  With this at least you will benefit from the ability to plot the changing states as a step time series.

If you would be open to having a chat with me about any other requirements in this product area, please let me know as I am actively gathering customer feedback in preparation for commencement of the work.

Kind Regards, Glen

Hi @jtfrandsen, we have had similar requirements in our project, to see for how much time a TS has been in which state. I’d like to bring three things to your attention which may be pertinent.

1. You can map string time series to integer time series using the synthetic time series expression function ‘map’.

   map(TS{externalId='stringstate'}, ['OPEN', 'CLOSED'], [1, 0], -1)

2. We use synthetic time series aggregated to the minute with step interpolation to get a dataframe which allows us to calculate how much time the machine spent in which status. You would need to use value_counts or something equivalent on the recieved dataframe.
3. To get the consumption from power (kWh from kW or m3h from m3), we can do it using the average function in synthetic time series, provided the input time-series has isStep == True. 

@Glen Sykes and @APSHANKAR Sagar thank you both so much for the time you put into these responses. Both are very insightful. I’m planning to pursue converting these to numeric time series, but am looking forward to the new features coming soon.

Glen, I will message you about setting up a time to chat.

Thanks again!