Draft hero image — replace alt text in Studio

Case

Castellum

Real estate

Intelligent indoor climate solution for real estate developer

Deep learningTime-series classificationTabular MLCost savingProcess efficiency

Modulai built a machine learning model that predicts meeting room temperatures 25 minutes into the future for one of the largest real estate developers in the Nordics. By combining sensor, climate system, and booking data, the solution enables energy savings and improved employee comfort compared to traditional HVAC control systems.

Outcome

  • 25 min

    Predicting into the future

  • 1,000

    Sensors deployed

    • Challenge

      A traditional HVAC control system only has access to current and recent temperature variations. An intelligent controller can draw on a far broader set of data, such as whether an area is likely to be occupied, when setting its output. The aim was to save energy and improve employee comfort.

    • Solution

      Data sources controlled and gathered by the client were used to build machine learning models that predict future meeting room temperatures. Two offices were fitted with nearly a thousand sensors measuring temperature, CO2 levels, sound, and more. Additional data from the climate system and room booking system was included. Various convolutional and recurrent neural network architectures were tested and benchmarked against traditional algorithms such as linear regression and tree ensemble methods.

    • Tools

      The project used a stack of relational and time-series databases for data storage and retrieval, along with Python, TensorFlow, and Scikit-learn for modeling.

    • Value created

      The final model architecture showed a significant increase in predictive power compared to existing systems.

    Learn more

    Related content