Self-configuring Event Detection for Electricity Disaggregation

摘要: Event-based non-intrusive load monitoring (NILM) is a major class of techniques that are used for electricity disaggregation in buildings. Event detection algorithms play a critical role in the performance of these techniques. There are different deterministic and probabilistic algorithms for detection of the events that commonly call for a number of parameters that need to be tuned as these algorithms are applied to a new dataset. The need for tuning could be a barrier to wide adoption of these algorithms. Therefore, in this study, we are proposing a selfconfiguring event detection framework that integrates conventional event detectors, automated clustering algorithms, and data-driven parameter optimization to enable selfconfiguration. The framework has been evaluated on real-world data and demonstrated promising results in high-quality detection of events.