Benefits of Water Consumption and Reactive Power in NILM

This technical report investigates the integration of water consumption data into eventless Non-Intrusive Load Monitoring (NILM) models and compares its effectiveness with reactive power data. The study outlines the objective of enhancing energy disaggregation by utilizing auxiliary sensing signals, which can improve the accuracy of appliance-level energy consumption analysis. The authors present empirical results indicating that while water data is beneficial for appliances directly associated with water usage, the inclusion of reactive power yields greater overall improvements. The report details a methodology for incorporating these data types into NILM models without necessitating significant architectural changes. It also highlights the systematic analysis conducted on real-world data to evaluate the performance of various model configurations. The findings suggest that combining both water and reactive power data leads to substantial reductions in disaggregation errors, thereby advancing the field of NILM and its applications in energy management.