Paper Conference

Proceedings of eSim 2020: 11th Conference of IBPSA-Canada


Anomaly detection in district operation systems

Gerrit Bode, Ansas Bogdan, Marc Baranski, Dirk Müller
RWTH Aachen University, E.ON Energy Research Center, Institute for Energy Efficient Buildings and Indoor Climate, Germany

Abstract: Modern buildings, as well as sites and even city districts, accumulate large amounts of data collected by monitoring and energy management systems. With the integration these data systems into more and more buildings and districts, the amount of data available to the operators of these districts increases dramatically. At a certain point, the check the plausibility of every single measurement and set point collected in the system must be supported by automation. In this paper, we present the application of an anomaly detection algorithm in a real-life district operation scenario. The algorithm is trained with synthetic sensor data to create representations of time series data. These representations are collected for every sensor during normal operations. If a newly calculated representation is too far from the existing data, an alarm is raised. We applied the algorithm to data collected in a district with multiple building types and energy systems in Munich. A failure event in a meeting room in one of the buildings is more closely investigated, and the algorithm sucessfully detects the problem. The developed system is able to be deployed in the application scenario with little configuration effort and provides valuable insights to the maintainable teams.