Dipl-Ing. Ole Ziessler already joined the EASD team in 2017. At that time, still as a student, he designed and completed both his project work entitled “Model-supported test of deployment planning methods in district heating networks” and his diploma thesis “Application of Machine Learning Methods on Building Monitoring Data”. Both papers dealt with very exciting topics. We were all the more pleased when the Verein zur Förderung der Ingenieurausbildung der Gebäude- und Energietechnik Dresden e. V. awarded Mr. Ziessler’s diploma thesis with the 2nd prize in 2019. The awarding of the 2019 promotional prizes took place during the 8th specialist symposium on 16.01.2020.

“Application of Machine Learning Methods on Building Monitoring Data” – A brief overview

Due to environmental, political and technological development, monitoring data acquisition within the building energy sector is becoming increasingly extensive. In this regard new data driven methods are gaining importance specifically for system analysis and optimization of such systems and the associated technical components. This thesis aims to investigate the relevance and applicability of Machine Learning (ML) in solving application specific challenges, using the monitoring data of a modern energy system of a building, as a case study. A ML based toolchain was developed consisting of three general research objectives, namely learning of sensor associations for system understanding, data validation and concrete system analysis. Individual ML pipelines, applying the general pipeline architecture, were developed utilizing Association Learning (AL), classification and regression base ML algorithms. Thorough evaluations of each showed that AL had limited applicability, whilst the classification and regression-based pipelines performed well, offering successful solutions to the associated objectives of data validation and system analysis. Concretely the Random Forest showed the best results, being able to learn underlying system dynamics, with potential in reducing manual effort in the process of system analysis. The results show that ML is applicable to building monitoring data, providing promising methods in terms of automatization and data-driven solutions to interdisciplinary challenges, whilst giving insight into general ML pipeline design.