Load Profile Modeling Using High-Frequency Appliance Measurements for Nonintrusive Load Monitoring

Many load disaggregation algorithms have been presented, capable of identifying individual appliances' operation within a buildings' electrical load profile. Publicly available datasets offer the potential to evaluate and compare these algorithms. Due to differing application areas, datasets show inconsistent characteristics, so exploitation of this potential is not straightforward. Aim of the presented work is to contribute towards solving this problem. We present a methodology for utilizing high-frequency appliance measurements for load profile modeling, while trying to keep modifications of the measured appliance features to a minimum. The voltage measurements were used to separate the appliances' events and states in current for each grid period, to be able to recombine appliance behavior in a user-defined manner. The modeled load profiles showed high accuracy compared to aggregated measurements. The presented methodology can be applied to various datasets and appliance measurements, to create standardized load profiles for the development, evaluation and comparison of disaggregation algorithms.