Automation of Threshold Values

Deep learning algorithms to define best threshold values are key components in this project.

Threshold values play an important role in the optimisation of running HVAC-systems and their surveillance concerning unexpected operating modes. MST as energy management provider and HSLU – T&A in cooperation with Empa will be working on deep learning algorithms to get automatised values based on historical data.


The following points are addressed within the project:

  • How can historic energy data be used to address both, energy optimisation and surveillance?
  • How can knowledge of historic data be used to optimise threshold values in an automated way?
  • How much CO2 can be saved using automated threshold values?

Initial situation.

MST provides an energy management system for more than 14.000 buildings. Due to its unique pool of historical data, automated threshold values shall be derived based on conventional statistical methods as well as using deep learning algorithms.

Outlook and next steps.

The project’s first steps have already been funded by an Innocheck. Lately, a subsequent application has been submitted to Innosuisse. Therefore, the project’s start is expected in autumn.


Dr. Thomas Schluck
Lucerne University of Applied Sciences and Arts, Institute of Building Technology and Energy IGE