Working with meter, BMS and sub-meter data was a big focus in the analytics space. Three main themes emerged during the study:
- How do we work with the data?
- Improvements on existing modelling methods.
- Successful case studies.
Working with data
Despite the prevalence of prediction algorithms there has historically been a scarcity of research addressing how to actually gather and tidy BMS data. An issue that is often glossed over is how difficult it is to actually obtain workable BMS data. Several papers addressed this by acknowledging the issue and presented ways to begin working with unwieldy data. The point was made that visualisation is a great starting point for understanding data. Buildings Alive research analyst Cameron Roach’s paper built on this by presenting a way to represent sensor time series data and sensor metadata in a low dimensional space for easy exploration. There was also a push to make data more available with some research focusing on generating datasets containing known faults to better assess FDD approaches.
Improving existing modelling methods
There is also growing interest in the use of data technology to improve building modelling. A few papers discussed the potential for model based predictive control built on sophisticated machine learning algorithms, such as reinforcement learning, as an alternative to traditional HVAC fault detection and control. One interesting example was of an energy model built with whole building electricity data that was able to guide grocery store owners to detect equipment failure and verify savings. Buildings Alive’s Hao Huang presented a novel energy demand prediction technology capable of addressing the most challenging aspect of energy modelling, achieving reliable prediction results under unprecedented conditions.
Successful case studies
There must always be a business case for any advanced analytics. Several examples of rules-based fault detection based on sub-metering were presented. Buildings Alive contributed to this discussion by exploring how improved forecast accuracy can improve peak demand warnings. Managing peaks not only reduces connection fees for buildings, but can also reduce stress on the grid – effectively acting as demand side management.
Conclusion
Over the five days almost 300 researchers and industry members presented at the conference. It became clear that energy savings will continue to be improved by advancing our analytics capabilities over the coming years. A continued focus on smart use of building data will be an important factor in our energy efficient future.