working on a similar project, would like to get in touch

Sorry if you feel this intrusive but i think we are working on very similar subjects. I tried to understand the boptest approach but not i cannot say i've really understood yet. Seems to be quite complex and i need more time to dive into it.

Anyway, since 2020 I work on a AI controller based on a nvidia jetson nano to command the heating circuits of tertiary buildings.

This controller completely bypasses all existing physical controllers : it is not acting through the web and is directly installed in boilers rooms. The embedded intelligence has to adapt to a great variety of circuit configurations and is trained through reinforcement learning, mostly via offline-policy methods such as double dqn, on a simplified simulator developped for the occasion :

https://github.com/Open-Building-Management/EnergyGym

The very first prototypes are already deployed on the field. It is quite promising and except from very extreme situations, the first neural network shows a real autoadapt ability.

Would you be interested to get in touch ?

On the github repo, most docs are in french, but i've got an academic summary/presentation of the research that i can quickly traduce .

Best

Alex

ibpsa / project1-boptest-gym

working on a similar project, would like to get in touch #117