Project in the field of IoT for building energy systems in cooperation with Indoorclima

The smart management of HVAC installations leads to energy savings of 5% to 20% A thermal inertia algorithm should indicate when to power on/off the HVAC system to reach desired temperature at the desired time. Inertia model developed in the past running with limitations: not considered external temperature and can't be trained on more than 2-3 months of data.

The objectives of this project are:

• Create a model for the prediction of thermal inertia during power on and power off.
• Improve error metrics by means of feature selection/engineering as compared with the model currently used
• Create a model which can be trained on 1 year data without negative impacting error metrics

The report can be seen here.
The modelling script here.

Current development

Objectives:

• Scale-up to all locations and integrate into production (Azure DevOps)
• Use forecasts of outer temperature to further improve the model
• Use dummy variables to consider the occupation level of buildings (big stores)

First iteration of current development can be seen here.