Building systems are collecting numerous data, which are stored, but are not automatically evaluated. However, the stored data is the key to efficient operation and to quick recognition of errors and inefficiencies. Currently, the evaluation at runtime is usually limited to launching alarms in case of sever errors. Usually, the collected and stored data is studied by the Facility Manager sporadically, who manually reads the data and uses tools like MS Excel for rough interpretation.
To ensure permanent monitoring of the logged data, automatic surveillance and interpretation should be installed. Thereby proper functioning as well as gained output can be permanently monitored. Inefficiencies will be reported timely.
This project will develop algorithms on the base of errors and inefficiencies that have been identified in advance. Those shall be then automatically reported. The algorithms span over a broad spectrum ranging from simple plausibility checks of system parameters to inductive statistics (correlation, analysis of variance, time series forecasting), hidden Markov models (HMM), self organizing maps (SOM), and conditional random fields (CRF).
Based on already available monitoring systems, common systems with solar heating and cooling components will be inspected and tested with the algorithms. A prerequisite is a data base of known errors, which will be generated by the experts from the operator of the installation. They will prepare a set of scenarios. The key to success of extrACT will be multiplicity of algorithms, meaning that they can detect the same category of errors independent of the actual installation in a similar fashion. This will dramatically reduce or even avoid the effort for continuous adaption of algorithms to new systems.
Aside of the usual alarming in case of detected errors, extrACT should also develop improvements for existing control strategies for automatic corrections of errors (“cleaning program”).
At the end of the project a library of algorithms is provided for systems in the area of sustainable heating and cooling, which are able to automatically recognize the sensors and start the monitoring and surveillance process.
Seven pilot installations will be available for data collection and validation of algorithms. Since extrACT will concentrate solely on data interpretation, the required measurement equipment and communication and storage systems are already available and are not part of the project. Only the interfaces have to be set up in a short work package at the start of the project.
G. Zucker, J. Malinao, U. Habib, T. Leber, A. Preisler, F. Judex; “Improving Energy Efficiency of Buildings Using Data Mining Technologies”,2014, 23rd IEEE International Symposium on Industrial Electronics, ISIE, Istanbul, Turkey (to be published) ISIE2014MalinaoZucker_2014-04-14