TSYS School's Yesem Peker and Students Publish New Study on Fault Detection in Smart Buildings

TSYS School computer scientist Yesem Kurt Peker has extended her research program examining fault detection in smart buildings with a new study in IEEE Access that leverages machine learning techniques to predict and classify faults in energy consumption, thus providing actionable insights to proactively mitigate them. In this study, Peker, her TSYS School students Shashank Sekhara, Akshith Nukala and McAndrew Okwei, and her colleagues from Mohammed VI Polytechnic University and US Ignite, Inc., propose a data engineering and machine learning framework deployed in the cloud to predict energy consumption across multiple building types. Their dataset includes hourly energy consumption, weather conditions and occupancy data, and results from outlier/fault detection to model prediction provide a detailed comparative analysis on methodologies for implementing a scalable and efficient framework for buildings’ energy management systems that is adaptable and can be seamlessly applied across a wide range of organizational infrastructures, highlighting its flexibility and relevance to diverse energy-optimization applications. Experimental results show that their framework, referred to as XGBoost, delivers the best forecasting accuracy across building types. Moreover, a Bollinger-Bands-based detector effectively captures abnormal fluctuations via deviations and volatility around the moving average, further supporting the effectiveness of their cloud-deployed framework.