Novel Algorithms for Filtering and Event Detection in Non-Intrusive Load Monitoring

The rise of the Internet of Things (IoT) and Industry 4.0 has brought about a connected ecosystem where instruments and devices are interlinked. Within this framework, an array of smart sensors and measurement units are present, resulting in a substantial amount of data. Consequently, data analysis has become a central focus across numerous research domains, including power systems. Non-Intrusive Load Monitoring (NILM) is a data-driven approach that aims to disaggregate total energy consumption into appliance-level information, thereby facilitating the balance between supply and demand in electrical grids. Notably, preprocessing assumes a vital role in data-driven methodologies. Firstly, preprocessing enhances the quality and accuracy of collected energy data, ensuring reliable inputs for subsequent analysis. Techniques such as noise reduction, outlier detection, and data cleaning eliminate unwanted anomalies, ultimately improving data integrity. Secondly, preprocessing enables feature extraction, facilitating the identification of distinctive patterns associated with individual appliances within the aggregated energy signal. This paper focuses on proposing novel preprocessing methods that encompass the aforementioned advantages, thereby contributing to the field of NILM.