Research Information System
Thesis Type: Postgraduate
Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Computer Engineering, Turkey
Approval Date: 2017
Student: OĞUZ CAN KARTAL
Supervisor: CEVAT ŞENER
Abstract:Efficient energy consumption is a trending topic nowadays, which has serious effects both environmentally and financially. Commercial and industrial buildings waste huge amounts of energy because of lack of integrated optimization systems. In this thesis, a big data analytics architecture for large-scale multi-tenant energy optimization systems is proposed, which is capable of doing various near-real time analyses on sensor data with the help of machine learning models created from old sensor data. In order to build big data analytics handling subsystem there are several steps during the flow of the sensor data. Raw data collected from the sensors in the field to the system is parsed and turned into meaningful data containing required features. This meaningful data is used for predicting the forth-coming energy consumption values. Prediction feature of the system is carried out with a machine learning model created from old sensor data. This meaningful data is also used for updating this machine learning model, to improve the accuracy and provide compatibility of model with live sensor data. Prediction and model update analyses are implemented on the streaming sensor data, without first storing it to a database or file system to provide near-real time feature of system. A very important feature of the system is scalability, which means adding new tenants or increasing the frequency of sensor data arrival is handled by system.