Feature evaluation for energy disaggregation

Thumbnail Image
Κουτρουμπίνα, Χριστίνα
Journal Title
Journal ISSN
Volume Title
The challenge of energy disaggregation focuses on the process of separating the total energy consumption of a household into the individual consumption of each appliance. The objective of this work is to identify an efficient and effective solution for energy disaggregation using a specific din-rail meter provided by Meazon S.A. This meter offers a comprehensive set of measurement data, making it a suitable choice for energy disaggregation solutions. The study focuses on evaluating the impact of different features on the performance of energy disaggregation and examining the results of selecting specific subsets of features. More specifically, this master thesis focuses on developing a solution to accurately detect the presence of a washing machine in a domestic setting by analyzing its energy consumption patterns. The method uses a unsupervised machine learning algorithm, DBSCAN, and is based on the cyclic behavior of the washing machine’s motor. The energy consumption data is collected by the DinRail 3-Phase Advanced WiFi meter provided by Meazon S.A., which records the consumption at regular intervals. The process involves detecting the cycles of the washing machine, extracting the motor consumption from the total consumption, and testing the accuracy of the algorithm with real-world data. The outcome is a reliable and efficient system for monitoring and optimizing energy consumption in a smart and economical manner.
Energy consumption, Machine learning, Unsupervised learning, Clustering, DBSCAN, NILM, DinRail 3-Phase Advanced WiFi meter, Washing machine