Although buildings are designed for occupants, evidence shows buildings are often operated inefficiently with regard to occupants’ preferences or building performance [i]. Insufficient indoor air quality affects our work productivity and can also have economic consequences [ii]. Balancing the goals of maintaining comfortable indoor conditions and reducing energy use requires careful trade-offs. That is within a sector responsible for around 40% of energy consumption and 36% of CO2 emissions in the EU (making them the single largest energy consumer in Europe). Buildings are also a large contributor to peak demand and plays an important role in the energy transition towards 100% clean energy. 

There's a wide gap between what building owners need in terms of automation and control and what existing siloed legacy systems (e.g., building management systems, lighting systems), are capable of [iii]. For instance, the siloed system prevents the exchange of information which creates major barriers to unlocking the potential of buildings, or, the silos are typically designed with only one type of stakeholder in mind. Even though siloed systems are getting smarter, it is rarely motivated to replace everything in the building to keep up with the latest, which motivates the usage of supervisory control and data acquisition architecture to complement existing BMS.

The complexity and nonlinearity of building energy systems in a single building have a high demand for modeling techniques. Building energy system approaches can be categorized into physics-based (white-box) modeling approaches, data-driven (black-box) modeling approaches, and those in between (gray-box model) [iv].

Data-driven model predictive control emerges as a promising method to address these challenges; to minimize the building energy consumption during a forecasting horizon while maintaining Indoor air quality. The model uses the past setpoint values, the forecasted weather conditions, and the zone temperature constraints to output control setpoints for the next control horizon.

Data-driven models outperform other approaches in terms of simplicity, automation, and development engineering cost. Unlike physics-based models with a high demand for domain knowledge and a long modeling period, the data-driven modeling process can be fully automated [iv]. An essential prerequisite for DPC is high-quality data set.


Contact us to

  • future proof your existing IT-architecture for Data-driven Smart buildings
  • implement cost-effective AI/ML for building performance optimization to other buildings up to portfolio level

David Hälleberg, Energy Solution Specialist at Helicon Technologies



[I] O’Brien W. et al. Introducing IEA EBC Annex 79: Key challenges and opportunities in the field of occupant-centric building design and operation, Building and Environment. 2020

[ii] Allen J. et al. Associations of Cognitive Function Scores with Carbon Dioxide, Ventilation, and Volatile Organic Compound Exposures in Office Workers: A Controlled Exposure Study of Green and Conventional Office Environments. 2015

[iii] Wen J. Building Fault Detection and Diagnostics. In: Baillieul J., Samad T. (eds) Encyclopedia of Systems and Control. 2020.

[iv] Zhang L. et al. A review of machine learning in building load prediction. Elsevier. 2021

Want to know more about how we can work together and launch a successful digital energy service?