Bachelor's thesis Paul-Werner Neißer-Deiters
Generating Standardized Identifiers from Metadata of Technical Building Equipment by Machine LearningCopyright: EBC
Reducing the energy demand of existing buildings requires advanced energy management systems. In order to apply energy management systems to existing buildings, it is necessary that the data of the building management system (BMS) are accessible and identifiable in a known format. BMS provide metadata that define its data points by text labeling. As there is currently no widely accepted standard, these labeling texts differ greatly, often even within buildings, as different installers use different formats. These circumstances make a manual transformation into a format that is readable by the advanced control system necessary. The manual transformation requires detailed knowledge of the BMS and a significant human effort. By applying of machine learning algorithms this effort can be reduced as repeating patterns can be transformed automatically. For this paper a browser-based tool was developed that enables the user to manually identify and transform metadata of suggested data points by labeling it after a standardized scheme and then using machine learning algorithms to predict other data points. To achieve best performance of the developed tool, different machine learning algorithms are compared by applying them to data sets of multiple BMS.