Heating system control using MPC

Traditional control of heating systems is based on temperature curves that translate outdoor temperatures to a supply temperature. This type of control does not contain feedback and cannot plan based on weather forecasts or dynamical comfort constraints, which gives rise to too high inner temperatures and energy waste. 

By using AI based technologies for controlling the heating systems it is possible to obtain more stable inner temperatures, reduce energy consumption, and to include dynamical comfort definitions.

The property industry had an archaic way of controlling district heating. This led to large variations and excessive temperatures in the apartments.
Fraunhofer-Chalmers Centre has developed a Model Predictive Control (MPC) framework which by utilizing historical data can learn the dynamics of buildings and control the heating system using both measured indoor temperatures and external signals and constraints.
The developed control systems is currently deployed and implemented for over 35 heating systems in Sweden and the number of customers and users are rapidly growing.

Upcoming Events

This Implemented (In Production!) has no planned upcoming events.


Engineering, Operations, Research & Development
Better Customer Experience, Better Quality, More Efficient, Saving Cost, Smarter Product or Service
Prediction, Optimization
Explainability and Interpretability, Machine Learning, Recurrent Neural Networks, Supervised
Sensor Data, Time series