Buildings account for large part of global energy consumption. Besides energy consumed due to normal operation, a large amount of energy can be wasted due to faults in buildings subsystems. Fault detection and diagnostics techniques aim to identify faults and prevent energy waste, but are often difficult to apply in practice. Data-driven methods, in particular, require an adequate amount of fault-free training data, which is rarely available. In this paper, we propose a method for anomaly detection that exploits consensus among multiple identical components. Even if some of the components are faulty, their aggregate behaviour is overall correct, and it can be used to train a data-driven model. We test our method on variable-air-volume units in an existing building, executing two experiments grouping the components according to ventilation unit, and according to room type. The two experiments identified the same set of anomalous components, i.e., their behaviour was different from the rest of the group in both cases, and this suggests that the anomaly was not due to wrong group assignment. The proposed method shows the potential of exploiting consensus among multiple identical systems to detect anomalous ones.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited