Title: Toward Real-World NILM: Analytical and Empirical Procedures for Assessing Comparability and Reproducibility

Data: 25 de Maio de 2023

Hora : 9H30

Link para Sessão de Zoom:  https://videoconf-colibri.zoom.us/j/94360329943?pwd=dkhTUGNQQVlyWURCdzhkSDhrRnozZz09

Thesis Abstract:

Reducing greenhouse gas emissions is necessary to reach sustainable energy distribution and mitigate global climate change. Non-intrusive Load Monitoring, or load disaggregation, has been widely considered for providing feedback and optimizing consumers' energy consumption. However, there is a general lack of guidance concerning the dataset aspects and the procedures in experiment design that mainly affect performance evaluation. Dataset aspects involve the standardization of measurement campaigns and the variability of appliance characteristics and usage patterns. The evaluation process comprises the choice of performance metrics and best practices in experiment design. Further research is essential for improving the comparability and reproducibility of NILM approaches. Consequently, NILM systems' deployment in the real world can be achieved at a larger scale. This document addresses the problem motivation and the research steps to be conducted throughout the Ph.D. thesis. At first, various dataset and evaluation factors that impact comparability and reproducibility are identified. Then, four research questions are proposed for improving NILM research on those topics. The methods to answer each research question throughout the Ph.D. thesis are also presented.